Home

JProbe User Guide

1. Summary Report This report contains the content of the Summary view Memory Pools Pool Used Mb Committed Mb Available Mb Maximum Mb Total 10 2 18 7 8 6 167 3 Heap Subtotal 1 4 5 2 3 8 66 7 Eden Space 0 7 0 9 0 2 4 2 Survivor Space 0 0 0 1 0 0 0 5 Tenured Gen 0 7 4 2 3 5 62 0 Non Heap Subtotal 8 8 13 6 4 8 100 7 Code Cache 1 0 1 0 0 0 33 6 Perm Gen 7 8 12 6 4 8 67 1 Garbage Collection Overhead 0 0 Normal GC Data Collector Count Time ms Cumulative Count Cumulative Time ms Copy 0 0 00 84 101 00 MarkSweepCompact 0 0 00 0 0 00 Threads Thread State Attach Listener RUNNABLE D3D Screen Updater WAITING Philosopher 1 BLOCKED on demos performance diners ChopStick 238a47 owned by Philosopher 2 Philosopher 2 BLOCKED on demos performance diners ChopStick 9d6065 owned by Philosopher 3 Philosopher 3 BLOCKED on demos performance diners ChopStick 1a918d5 owned by Philosopher 4 Philosopher 4 BLOCKED on demos performance diners ChopStick 10ea9ba owned by Philosopher 5 Philosopher 5 BLOCKED on demos performance diners ChopStick 1bdc9d8 owned by Philosopher 1 346 JProbe User Guide Recorded Level Full Encapsulation Recorded Time 19s Timers Elapsed CPU Elapsed Time of All Threads Combined A 3m CPU Time of All Threads Combined B 437ms CPU Utilization Rating 0 0 Methods to Investigate Method Method Elapsed Time of B java io PrintStream printin java lang String 10 7 java awt Graphics d
2. sssssecsesssessssssssstecnssssessssnsessseesesstsssecnsessessvenens 460 460 JProbe User Guide Starting and Viewing a Remote Session The following procedure assumes that you have a session settings JPL file and a startup script BAT CMD SH file that were either created on your local machine and uploaded to the server or created on the server machine using the JProbe Console Note If you created the JPL file on your local machine ensure that all server paths are correct for the remote server environment If you are analyzing a Java EE application on an application server this procedure assumes that you have created an application server integration on the server For more information on using the command line to start Java SE and Java EE applications see the JProbe Reference Guide To start and view a remote session 1 Connect to the server via a command window 2 Run the startup script that you created in the JProbe Console JProbe starts your program on your application server You should see a message like the following JProbe Connection Manager is available for Console connections on port 52991 Note The default port number is 52991 If this port is unavailable or if you have specified a different port in your JPL file arguments or on the command line the port number is different 3 Start JProbe on your local workstation 4 Click Attach to Session on the toolbar The Attach to Running Session di
3. At the end of this process you will have a settings file and a startup script Note If you are running JProbe in a heterogeneous networked environment for example UNIX on one computer and Windows on another configure JProbe on the computer that runs your application This ensures that the correct syntax for the operating system is used To set up JProbe to analyze a Java SE application 1 Launch the JProbe Console on the computer where you run your application Windows Click Start gt Programs gt JProbe gt JProbe Console UNIX Navigate to JPROBE_HOME bin and execute jpconsole 2 Click Configurations in the Navigation area 28 JProbe User Guide Right click Java Application and click New Configuration or Select Java Application and click Configurations gt New Configuration Tip Alternatively you can use the standalone version of the Configuration tool called jjpcontig You can launch jpconfig from the Start gt Programs gt JProbe gt JProbe Config menu in Windows or from the JPROBE_HOME bin directory In a headless environment you can use the command line version of joconfig For detailed instructions see the JProbe Reference Guide The New Configuration Wizard dialog box appears Tip If your application server is not in the list you may still be able to use JProbe For more information see What Can Do if My Application Server Is Not Supported on page 64 Provide a name for this config
4. Icon Name Type Root Root instance 1 Static Root Static root instance E fe Recording Instance created at the recording boundary Boundary Cycle Instance that refers to itself circular reference Note The status bar also displays the most common icons and short descriptions of what they represent Understanding the Leak Doctor The purpose of the Leak Doctor is to model what would happen if you removed a reference You can expand collapse nodes in the tree and you can remove edges After removing references you can restore them if you like The Leak Doctor starts at the loiterer and traces the paths back through instances that reference the loitering object using a breadth first algorithm In other words it looks at all the children of a referrer for objects of interest before moving on to its grandchildren The search stops when a referrer is discovered that has a comment The comments are explained in the following table Comment Indicates From Class A static field which can remain in the heap for a long time possibly housing a problematic collection Exploring Memory Analysis Views 193 Exploring the Leak Doctor View Comment Indicates From Root A root that is responsible for pinning an instance see a description of root types below the table Recording Boundary A reference from an object that has been around for a long time to an instance that has been around for a shor
5. Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables and the same command line arguments Usage lt configuration_startup_script gt Note Windows XP and Windows Vista can be used to develop applications but are not supported by Oracle as production deployment platforms Starting Your Application with JProbe Attached 93 Runtime Notes for Application Servers Oracle OC4J 10g v10 0 x You run the JProbe generated startup script using the following syntax lt configuration_startup_script gt Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Oracle OC4J 10g v10 1 2 or below You run the JProbe generated startup script using the following syntax lt configuration_startup_script gt Important It is recommended that you use the default script directory that is the script d
6. exclude v v V demos gamepack framework image GameImage exclude v v To generate filters automatically based on the items in the Elements table click Create Filters The reale ters J new filters are appended to the current list of filters Elements Enter the war ear and or jar files containing the code of your application Element C Program Files ApplicationServers BEA_Systems user_projects applications gamepack gamepack ear Insert Delete My Application Filters Table If you are configuring a Java Application you specified a main class for your application By default JProbe adds the package containing that class to the My Application Filters table For more information see Main Class and Working Directory on page 32 If your Java Application requires other custom packages or if you are defining a Java EE application you can add your custom packages to the My Application Filters table If you want to exclude some of the classes or methods in a package you can enter additional filters for this code below the package filter and set the action to Exclude 40 JProbe User Guide Filters positioned lower in the table override filters on the same code positioned higher in the table Note When you exclude from your package the code that you don t need to analyze JProbe runs faster and records lower overhead during the data collection Inthe case of a memory analy
7. 4 Click OK when you have finished adding filters The view refreshes and displays only the instances that meet the filter criteria you have set The bar chart is redrawn to reflect the matched instances Syntax for Instance Filters Filters are case sensitive and blank spaces are not allowed You can use an asterisk as a wild card it matches any character including the package separator If you specify the code in a shorthand or partial form JProbe automatically changes it to canonical form The change may not be exactly what you desire so be sure to verify it JProbe does not check to see whether the filter actually matches anything in your code For Exploring Memory Analysis Views 143 Exploring the Instances View example if you make a typing error in your package name nothing in your code will match the filter and so the filter is not effective The following table contains common syntax and describes how JProbe interprets it If you type this M P String P C do Pre Canonical form is this ke oh eC EC aTi C SMC P String no change Pre And filters are applied to All methods in all classes in all packages including the unnamed package All methods in all classes in the unnamed package only All methods in classes named C in any package All methods in the class named C in the unnamed package only Method in all classes in all packages
8. Required Field Java Executable Use a JVM that reflects the JDK environment under which your program was developed and that is also supported by JProbe JProbe supports the most popular JVMs however there may be some known restrictions For a list of supported JVMs and known issues see the JProbe Installation Guide Configuring JProbe to Analyze Your Application 37 Specifying a Java Virtual Machine To select a JVM 1 Click the browse button beside the Java Executable field 2 If your JVM is listed select its check box 3 Otherwise click an empty cell and do one of the following e Click the cell browse button navigate to the Java executable and click Open e Double click the cell and enter the fully qualified path to the Java executable The check box for this JVM is automatically selected Java Virtual Machines JMs C Program Files JProbe 8 1 bin jre bin java exe C Program Files JProbe 8 0 1 bin jre bin java exe C Program Files Java jdk1 6 0_04 bin java exe v 4 Click OK Java Options You can add Java options as you would on the command line To specify Java options 1 Click the browse button beside the Java Options text box 2 Do one of the following e Enter the options in the text box as you would on the command line and click Parse Arguments The table is populated with a list of options If an option was parsed incorrectly you can double click
9. Task Examine the methods in a class Action Choose the class in the upper table Response The lower table displays a list of methods in the chosen class Show only the methods with a percentage of hits less than a specified threshold Set Show Only Methods with Hit Lines to a percentage of hit lines Methods with Hit Lines less than the chosen percentage are shown The rest of the methods are hidden Sort a table by a metric Click the column heading for the metric The table is sorted To reverse the sort order click the heading again Show hide columns Right click and choose Show Hide Columns Choose the columns to show and click OK The chosen columns are displayed The rest are hidden Rearrange columns Drag a column by its column heading and drop it in a new location The column appears in its new location Exploring Coverage Analysis Views 377 Exploring the Coverage Session Task Resize column widths Action e To set the width drag the edge of the column heading left or right e To size a column to fit its data right click the column and choose Size lt column gt to Fit e To size all columns right click and choose Size All to Fit Response The affected columns are resized When sizing all columns to fit if the resized columns do not fill the table the width of all columns is increased to fill the empty space Expo
10. Task Show hide all columns displaying CPU Times Action select a node in the Method List right click and select Show All CPU Columns Response All columns displaying CPU Times are displayed hidden if this information is available Hide the column currently selected Hide Column lt column name gt The column currently selected that is the column based on which the Call Tree table is sorted is hidden Resize column widths To set the width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Sort the table by a metric Click the column heading for the metric The table is sorted To reverse the sort order click the heading again Editing the Data Set The data set in the snapshot reflects the data collected during the session You can temporarily remove data from the current data set by pruning some trees or isolating a single tree All the calculated metrics are automatically recalculated for the data that remains JProbe keeps a history of all edits to the data set that is separate from the history it keeps for hide show actions You can revert to the entire data set by resetting the graph For more information see the following topics e Isolating
11. The Instance Detail view opens with all the instances allocated by this method in the Instance List View the source code for an allocating method Select a method right click and select Show Allocated At Source Note _ If this menu item is disabled JProbe cannot associate a source file with the class file The Source view opens with the source code for the allocating method You may be prompted to locate the source code file Drill down into a category node Select the node and click Replace Category with More Detail 12 The node expands to one or more nodes Remove a node from the tree Select the node and click Prune te The path starting at the selected node is removed from the tree Focus on a single path Select the node and click Isolate i Only the paths through the selected node are displayed Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden 184 JProbe User Guide Task Resize column widths Action To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit Res
12. cc cciissesisiecasasiscienaessaatidssasnsnicssiaienateatitiees 454 Changing How JProbe Exports toa OSV Finnsson aa 455 Hiding Package Names and Method Signatures cecessesssesssseeseessestesneetesneetesnseneneeatenees 455 Specifying Source File LOCAtONS ssesessesesctestessesneeeeeneesteneeateneesteseeatesnseateneeateneeatenseatenees 455 Configuring Memory Pools 454 JProbe User Guide Controlling How JProbe Operates This section describes how to change the default operating settings such as hiding the welcome pages and disabling warnings To change how JProbe operates 1 Click Options E4 on the toolbar or select Tools gt Options from the menu bar By default General options are displayed 2 Do any of the following To show or hide the Welcome pages select or clear the Show Welcome Pages check box To turn ON OFF the JProbe warning about unsaved snapshots when you exit select or clear the Warn on Unsaved Snapshots check box Caution When this options is disabled temporary snapshots are deleted when JProbe exits To turn ON OFF the JProbe warning about comparing snapshots from different applications select or clear the Warn when Baseline and Snapshot to Compare are from Different Applications check box To change where the JProbe Console stores the temporary files it requires while operating specify the fully qualified path to the directory in the Directory field Note T
13. public String createFile setupFile String filename createFileName Would like to take a snapshot here if filename is null writeHeader filename You want to take a snapshot before writeHeader is called but only if filename is null You would add the following method private void triggerPoint And change createFile as follows public String createFile setupFile String filename createFileName 432 JProbe User Guide if filename null Would like to take a snapshot here if using JProbe and filename is null triggerPoint writeHeader filename It is at this point that you set a trigger on entry to the triggerPoint method using the procedure in Adding Triggers on page 425 Creating and Managing Categories Categories provide a powerful tool for organizing the data that JProbe collects about your application JProbe provides some preset categories such as Java SE and Java EE but you can also create your own By default JProbe creates a category for the code that you specify in the My Code tab of the Create Edit Settings dialog box Categories are used in the Memory analysis tool to group instances together Adding Categories You add code to categories using filters For information about filters see Syntax for Filters on page 434 To add a category 1 Select Tools gt Manage Categories Refining Data Collection and Presentation 433
14. Coverage Session Report All Coverage Session reports include a report header and a summary of the classes in the Coverage Session view You can also choose to include a summary of the methods Create Coverage Session Report Output File iLocal Settings Temp Adventure_TestCase1 html Bi Report Content Class Summary O Class and Method Summaries Generate File Cancel Cicin J Help You can create the following reports e Class Summary e Class and Method Summaries Class Summary The Class Summary contains a list of classes in the runtime view and the percentage of methods that were hit by the test suite Classes Name Methods Hit demos gamepack ejb user UserBean 16 7 demos gamepack framework serviet GlobalServiet 90 0 demos gamepack ejb game GameHandlerBean 14 3 demos gamepack fremework session HttpGemeSessionManager 11 1 demos gamepack framework serviet LoginitServiet 75 0 demos gamepack framework util Xmlutil 100 0 demos gamepack framework GameGlobal 33 3 demos gamepack framework database JDBCDatabaseManager 11 8 demos gamepack framework config GamePackController 714 demos gamepack framework config ManagerProperty 100 0 demos gamepack framework config GamePackGame 58 3 demos gamepack ejb game GameBean 14 3 demos gamepack framework user EJBUserManager 14 3 demos gamepack fremework config GamePackManager 87 5 demos gamepack framework config GameMode 71 4 demos gamepack framework user EJBUserM
15. The View area also displays a Configuration Details tab which guides you through the process of configuring JProbe to analyze a Java SE application or a Java EE application server that is currently not defined When you select a configuration in the Configurations Navigator the View area displays a lt Configuration_File_Name jpI gt tab This tab allows you to edit the settings for the selected configuration create a JPL file and a script file to integrate your application with JProbe run your script file or run your script file and auto connect your application with JProbe Note JProbe demo configurations delivered with the product cannot be modified or deleted Snapshots Navigator The Snapshots Navigator displays snapshots taken during analysis sessions You can also load previously saved snapshots The View area by default displays the Welcome to JProbe tab This tab contains three tabs across the bottom one for each of the analysis types Memory Performance and Coverage and includes Quick Start links shortcuts to JProbe community page and to documentation The View area displays the Welcome to JProbe tab until you connect to a JProbe session When the JProbe Console connects the current analysis tab updates to match the analysis type selected in configuration It then displays the runtime view for that analysis type You can interact with the runtime view for example you can update the runtime view to see the data c
16. V Enable JDBC component instrumentation m select JDBC components for instrumentation 7 com solarmetric config a db2j config 7 MZ com sun config 7 oracle config al Timing Settings for Performance You can select the Timing settings for your Performance analysis in the General tab of the configuration wizard By default JProbe calculates method execution times using elapsed time Elapsed time includes program pauses such as wait times for resources By including pauses you gain an understanding of the performance of your program from a user s perspective You may also find it helpful to judge the performance of your code separately from the performance of the environment in which it runs In this case you can omit program pauses by selecting CPU time CPU timing is not available on all operating systems Learning More About Performance Options 267 Analysis Options for Performance Timing Description Elapsed Default Elapsed time includes program pauses such as time spent waiting for I O CPU CPU time ignores pauses JDBC Component Instrumentation You can select the JDBC Component Instrumentation settings for your Performance analysis in the General tab of the configuration wizard By default all JDBC components identified for instrumentation are enabled Database component instrumentation is offered by identifying driver classes for instrumentation This
17. a Both of the above Note In the Total row the total Child CPU and Child Elap data matches the information displayed in the Summary pane CPU Child Time and Child Time respectively Calling Context Sites The Calling Context Sites CCS table shows individual CCT nodes For more information about the CCT see Understanding Performance Data on page 222 It allows you to inspect the CCT directly instead of viewing data through the Call Graph The action of clicking a parent or child from the edges tables populates the CCS table Exploring Performance Analysis Views 331 Exploring the Method Detail View with CCT nodes that have a parent or child The CCS is a tree table with child nodes that are collapsed Click the to expand the children Root main 6 1 A 4 2 B 9 3 X1 3 2 X2 3 2 C 4 2 X3 2 2 l X4 1 1 x5 2 1 X1 2 2 4 D 1 1 Simple CCT For example the Call Graph can only show you aggregated data for x1 But if you view the method in the Method Detail view we can see the separate X1 nodes if we select in the Edges From the Parents table the Calling Context Sites table shows a single row with 2 calls and method time of 3 and we can see it called x4 This is the left most node under A If we select C from Parents we see the other node showing 2 calls and a method time of 2 and we can see that it never call
18. c Double click the method you want Refining Data Collection and Presentation 427 Setting Triggers e Enter the package class and method no signature in the cell Wild cards cannot be used to specify the whole method name For example x x methodD and myPackage MyClass set are valid but myPackage MyClass or are not valid The checkmark at the beginning of the row indicates that the trigger is enabled A default event and action are applied to the method 3 To set when the trigger fires select an option from the Event list 4 To set the action to occur when the trigger fires select an option from the Action list For more information about each action see e Triggers for Memory on page 125 e Triggers for Performance on page 271 e Triggers for Coverage on page 371 5 Change the options in the lower part of the screen if desired e For Memory analyses see Setting the Recording Level on page 110 e For Performance analyses see Setting the Recording Level on page 236 e For Coverage analyses select Data Recording Off or Data Recording On 6 Click Save If you want to refine the action of the trigger you can click Advanced Control to open the Set Advanced Trigger Control dialog box For details see Setting Advanced Trigger Controls on page 427 Setting Advanced Trigger Controls The Advanced Trigger Controls options allow you to have more control over when and or how
19. org w3c org xml sun com jprobe Note An error message is displayed during the integration process when no com jprobe filter strings exist in the java6 server profile file You run the JProbe generated startup script in the same way that you run your startup script dmk bat or dmk sh JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly What to Do When the Console Does Not Connect If the JProbe Console does not connect to an application running under JProbe you can check to see if any of the following issues apply to your situation Port Number Mismatch The JProbe Console may be listening on the wrong port In the JProbe Execution Console window the window that opened when you ran the JProbe generated startup script note the port number that the JProbe Analysis Engine is using for this session Click Attach to Session on the toolbar and ensure that the same port number is specified in the Port field for the selected JProbe Analysis Engine Alternatively check the application server log files in case y
20. Advanced JProbe Options Enter Arguments As Command Line Text 3p_output_file gamepack Parse Arguments Enter Arguments As One Per Line jp_output_file gamepack A 3 To re arrange or delete options from the table click a row and use the buttons provided 4 Click OK If there is a mismatch between the contents of the text box and the table JProbe prompts you to select a set of options to use The default is the table The options are listed in the JProbe Options text box JProbe Port JProbe uses the first available port number starting at 52991 If the port is unavailable it tries 52992 52993 and so on When you run the JProbe generated startup script for your application the actual port number used for JProbe is displayed in the JProbe Execution Console window If you want to specify a different port number enter that value in the JProbe Port field Note When you connect the JProbe Console you may need to update the port number in the Attach to Running Session dialog box to match the port number displayed in the JProbe Execution Console window if the JProbe Analysis Engine is not already displayed with the correct port in the list Snapshot Basename When JProbe takes a snapshot it automatically assigns the name snapshot to the first one and adds an incremental number to subsequent snapshots snapshot_1 snapshot_2 Configuring JProbe to Analyze Your Application 49 Reviewing t
21. Elapsed recursive time CPU Recursive Time CPU recursive time Graph The graph area displays the method that you are interested in in red and also up to five parent methods located to the left of the current method and associated child methods located to the right of the current method Detailed information is shown in the Edges From Parents and Edges To Children panes Edges From Parents The Edges From Parents pane displays information about the current method that you are investigating with respect to each of its parents This information is directly related to the information held in the CCT The parents contribute to the overall value of the method that you are investigating Each row shows the contribution of a particular parent The data corresponds to what you might see if you hovered over the edge from the named parent to the current method in the Methods view Exploring Performance Analysis Views 329 Exploring the Method Detail View The following table lists information presented in the Edges From Parents table Column Data Method Description The name of the parent method that called the current method This column may contain recursion indicators t A recursive indicator can appear in this column if the parent is a recursive caller that is if it is the parent of one or more recursive nodes Cumulative time is not included in recursive nodes Calls Displays the number of times that
22. View only directly Click Show Directly The table is updated to allocated objects reflect the selected option Allocated Instances View all allocated objects directly and indirectly Click Show All Allocated uh Instances 22 The table is updated to reflect the selected option Examine the instances for the allocated object Select a class and click Instance Detail View E OR Select a class right click and select the Open Instance Detail View option from the list The Instance Detail view opens with all the instances for the selected class listed in the upper pane For more information see Exploring the Instance Detail View on page 166 Examine merged stack traces Select a class and click Merged Allocation Points View Be OR Select a class right click and select the Open Merged Allocation Points View option from the list The Merged Allocation Points view opens and displays the merged stack trace for the class selected from the Allocation Detail view To follow a stack trace click the plus sign to expand the list For more information see Exploring the Merged Allocation Points View on page 181 Exploring Memory Analysis Views 165 Exploring the Allocations Detail View Examine the allocation call tree and graph Select a class and click Call Traces View EX OR Select a class right click and select the Open Call Traces View option fr
23. on page 455 Exploring the Instances View The Instances view is the starting point for investigating a heap It displays data for the entire Java heap and contains data on the instances of classes objects allocated during the execution of your program You can apply filters to see a subset of the data This section contains the following information e Opening the Instances View on page 138 e Understanding the Instances View on page 139 e Filtering Data in the Instances View on page 141 e Interacting with the Instances View on page 144 Opening the Instances View The snapshots that you took during a session are listed in the Snapshots Navigator To open the Instances view 1 Inthe Snapshots Navigator select the snapshot that you want to view 2 Double click the snapshot Exploring Memory Analysis Views 139 Exploring the Instances View Understanding the Instances View The Instances view consists of a horizontal bar chart and a table welcome G Runtime Summary BE instances 7 Vlae RK Leg Oe Investigate by Heap Count Click on chart segments or legend items to drildown on categories Categories M Java Base Libraries DPrimitive Arrays Bava UI Toolkits Bother Java SE sun Other Filter Classes 161976197 Name Recorded Count Heap Count Recorded Memory Heap Memory Keep Alive Size Dead Co
24. By default Jprobe accumulates data for the methods in your program This level of detail is sufficient for finding performance bottlenecks After you find a bottleneck however you may want more information on the methods contributing to the bottleneck You can rerun the session with a filter on the problem method to track time for each line of code in that method The metrics collected are similar to the method level metrics but are adjusted to represent lines of code Cumulative and Averages metrics are also for each line To see line data you need to open the Source view Note Collecting line data is time consuming and will impact session performance and the accuracy of other method data Use Line Level detail only after you have found one or two methods that contribute to a performance bottleneck To view line level data for methods you must have the corresponding source code With line level detail JProbe collects data for each line listed in your compiled program s line data table As you may know the line table matches up the compiled bytecode with the actual line of Java code in your uncompiled program Sometimes the match is imprecise because the bytecode represents more than one line of code For the purpose of debugging performance bottlenecks however the line data table provides enough accuracy to point you in the right direction You can try to improve the line data table by compiling your program with the Java compiler
25. Cancel Refresh 4 Select an engine from the list of currently running engines and click OK The JProbe Console begins receiving data collected by the JProbe Analysis Engine The Coverage tab is selected and the Coverage Session view is displayed Running a Coverage Analysis 363 Investigating Your Results Welcome F Coverage Session Gj Snapshot Browser 7G Qe 2a class Hit Methods age adventure Room Ej demos coverage adventure Adventure 7 demos coverage adventure Logic 7 demos coverage adventure Roomlist 200 Show Only Methods with Hit Lines lt 10 v Method Hit Hit Lines Method der enture lt init gt demos coverage adventure Room lt init gt demo demos coverage adventure Room checkSelecti Yes 100 0 demos coverage adventure Room setAssociate No 0 0 demos coverage adventure Room setDescripti No 0 0 demos coverage adventure Room setDirection No 0 0 demos coverage adventure Room setNametint No 0 0 demos coverage adventure Room setNumDirec No 0 0 demos coverage adventure Room showDescrip Yes 100 0 demos coverage adventure Room showDirectio Yes 100 0 Session Connected Finishing current recording and taking snapshot Snapshot C Quest_Software JProbe_8 3 snapshot_raw_S jpc successfully written Snapshot C Quest_Software JProbe_6 3 snapshot_raw_5 jpc has been added to th
26. Click Run amp Connect The script is launched automatically the JProbe Console is Configuring JProbe to Analyze Your Application 55 Saving the Configuration and Integrating automatically attached to your running application and you can start to analyze your application right away For more information see Running Your Script and Connecting the Console on page 58 Saving the Configuration and Integrating When you click Save amp Integrate JProbe first saves the configuration settings see Saving the Configuration on page 50 then automatically launches the Integration wizard Integrating Your Application on page 51 Running Your Script If during the integration stage see Executing the JProbe Startup Script on page 53 you have selected to run the JProbe startup script the script is launched automatically and you can start to analyze your application right away If you want to specify any additional arguments to the script file or change the working directory you can use the Run button which opens the Run dialog Note Before running the JProbe startup script you must manually set any environment variable required by your application You can run any JProbe script manually using the JProbe Console or the JProbe Configuration tool You can also run a JProbe startup script using the command line tools from a command window Running Your Script from JProbe Console To manually run your JProbe
27. Dead Count Memory Select a class from the table and click Merged Allocation Points Be Note This action is available only if traces were collected for the selected classes The Merged Allocation Points view opens with the merged stack traces for the selected classes For more information see Exploring the Merged Allocation Points View on page 181 See a call tree graph for selected instances or garbage data Set Investigate By to Recorded Count Memory or Dead Count Memory Select a class from the table and click Call Traces BK 7 Note This action is available only if traces were collected for the selected classes The Call Traces view opens showing the allocation call tree and graph for the selected instance s For more information see Exploring the Call Traces View on page 174 146 JProbe User Guide Task Show allocations of large number of objects Action Select a class from the table and click Show Allocations View E Note This action is available only if trace information was collected Response The Allocations view opens showing all the methods that allocated instances For more information see Exploring the Allocations View on page 160 Drill down into the data Click a segment of the bar chart The Instances view is redrawn to display only the instances in the chosen category Filter the table See Filtering Da
28. JProbe can also use this information to provide default data collection filters This means that your analysis collects only the data you are interested in For more information see Refining Data Collection Filters on page 422 Configuring JProbe to Analyze Your Application 39 Specifying Your Code You specify your custom code using application filters JProbe may propose some default application filters based on information that you provide elsewhere in the configuration such as the main class for a Java Application You can enter other packages one per row and or you can have JProbe create filters for you from elements such as a JAR file WAR file or EAR file B New Configuration Wizard Specify Your Code Optional Tell us about your code using application filters JProbe creates a category containing the code that you set to Include A category is particularly powerful in a Memory analysis where you can click the category to see only instances created by your code For Coverage and Performance analyses JProbe uses application filters as default data collection filters You can set or change data collection filters later on Category Program Name Demo To create an application filter specify a package class or method and select an action My Application Filters package class method Action demos gamepack framework Game include demos gamepack framework faults Deadlock
29. JProbe keeps a history of your hide and show actions as graph views As in a Web browser you can go back and forth in the view history using the Back and Forward toolbar buttons Customizing the Call Graph You can change how time is displayed in the method list Tools gt Edit Options gt Data Display gt Performance You can also customize your view of the Methods view by changing the font family size and style as well as the color scheme e Changing the Scale Gradation e Changing the Color Scheme e Customizing Fonts Changing the Scale Gradation By default the scale gradation is geometric which means that except for the first two ranges the range of the segments doubles from minimum to maximum For example if the first segment spans 15 units the second spans 15 the third 30 the fourth 60 and so on doubling in size until the maximum value is reached Use a geometric gradation to see the relative relationships among expensive nodes in the graph in particular when a graph has a few nodes with values that far exceed the other nodes There will be times however when you want a quantitative view of the graph In this case you can change the scale gradation to linear the color scale is divided into segments of equal size and the graph highlights the few most expensive nodes To select the gradation used for the scale e To highlight quantitative differences among the nodes right click the color scale and select L
30. Pool button You can choose a different time interval for the summary graph from the Pools and GC Interval list The options are One Minute Five Minutes Fifteen Minutes Thirty Minutes and One Hour The data that is displayed in the Memory Pools graph varies depending on the JVM and the options that you are using For JVMs that use generational garbage collectors heap pools typically represent the generations For example in Sun JVMs with some options as shown above Tenured Gen is the old generation while Eden Space and Survivor Space together make up the new generation 134 JProbe User Guide The memory usage of the pool is an estimation provided by the JVM It consists of three values which appear in tooltips when you hover over a gauge Used The amount of memory currently being used Committed The amount of memory that the OS has guaranteed to be available to the JVM This will always be greater than or equal to Used Max The maximum amount of memory that pool will use Not all pools have a maximum value Maximum value is often affected or directly set by JVM options like xmx A maximum value will always be greater than or equal to the committed value JProbe defines Available Memory as the difference between committed and used GC Data Tab The GC Data tab contains a graph and a table The graph displays each type of garbage collection at the time it occurred The Count ticks at the right of the graph indicates how
31. Quest Central Quest vToolkit Quest vWorkSpace ReportADmin RestoreADmin SelfServiceADmin SharePlex Sitraka SmartAlarm Spotlight SQL LiteSpeed SQL Navigator SQL Watch SQLab Stat StealthCollect Storage Horizon Tag and Follow Toad T O A D Toad World vAutomator vControl vConverter vFoglight vOptimizer Pro vPackager vRanger vRanger Pro vSpotlight vStream vToad Vintela Virtual DBA VizionCore Vizioncore vAutomation Suite Vizioncore vBackup Vizioncore vEssentials Vizioncore vMigrator Vizioncore vReplicator Vizioncore vTraffic Vizioncore vWorkflow WebDefender Webthority Xaffire and XRT are trademarks and registered trademarks of Quest Software Inc in the United States of America and other countries Other trademarks and registered trademarks used in this guide are property of their respective owners Disclaimer The information in this document is provided in connection with Quest products No license express or implied by estoppel or otherwise to any intellectual property right is granted by this document or in connection with the sale of Quest products EXCEPT AS SET FORTH IN QUEST S TERMS AND CONDITIONS AS SPECIFIED IN THE LICENSE AGREEMENT FOR THIS PRODUCT QUEST ASSUMES NO LIABILITY WHATSOEVER AND DISCLAIMS ANY EXPRESS IMPLIED OR STATUTORY WARRANTY RELATING TO ITS PRODUCTS INCLUDING BUT NOT LIMITED TO THE IMPLIED WARRANTY OF MERCHANTABILITY FITNESS FOR A PARTICULAR PURPOSE OR NON INFRINGEMENT IN NO EVENT
32. Tip The computer may be localhost or a remote computer The port number must be the one specified in the JProbe Execution Console window d Click OK The JProbe Console connects to the JProbe Analysis Engine or to the JVM that is running with the application The runtime view for the analysis type is displayed 3 To learn more about monitoring and interacting with sessions see the Running an Analysis topic for the analysis type that you chose Running a Java EE Application with JProbe When you integrated your configuration JProbe created a startup script for your application server This is the file that you use to run your application under JProbe For more information see Integrating Your Application on page 51 If you are running a remote session you need to have JProbe installed on both computers Your license provides for this double installation To run your Java EE application with JProbe 1 Important Review the runtime notes for your application server See Runtime Notes for Application Servers on page 75 2 Ifthe runtime notes indicate that an application server configuration file will be edited and later restored make a backup copy of the file 3 On the computer where you run your application run the JProbe generated startup script from the command line or from the JProbe Console s Run dialog The syntax is in the runtime notes for your application server The JProbe Execution Console window is
33. efforts on frequently attempted use cases you can ensure that your improvements will be visible to your users To help ensure that performance gains are visible 1 Identify the use cases for your program 2 Analyze which use cases are attempted most frequently that is determine the critical path of your program Learning How to Use JProbe Effectively 415 Hints for Success with JProbe 3 Identify the use cases in the critical path that are prompting complaints You will focus on this subset of use cases first Formulate hypotheses for the use cases Run a Memory analysis on the use cases 4 5 6 Runa Performance analysis on the use cases 7 Run a Coverage analysis with the test cases that target the use cases For example in a web based application a high level task might be to log on to a secure area with a user ID and password This case implies a set of steps that your user goes through such as obtaining a unique ID setting a password entering the data into an online form submitting the form for validation and either going to the secure site or being denied access If the validation step is a perceived problem you could run Memory and Performance analysis sessions on this area only With Coverage you can ensure that the entire process is tested Compiling Your Program You can run JProbe without recompiling your program for most situations If you intend to analyze your program line by line however you may want to c
34. frequent steep spikes in the Memory Pools chart To determine the source of the allocations 1 Inthe Runtime Summary view click Data Recording 2 The Set Recording Level dialog box appears 2 Select Record Allocations Stack Traces and Garbage Data For All Instances and click OK 3 Exercise a use case in your application to generate activity in the JVM heap 4 Click the Heap Data tab of the Runtime Summary view and click Refresh Runtime Data Eea It may take several seconds for data to be collected from the JVM 120 JProbe User Guide 5 When the table has been populated with data select Dead Count or Dead Memory from the Investigate by list The bar at the top of the view shows a breakdown of the objects that have been allocated and garbage collected during data recording You can drill down by clicking on segments of the bar or you can sort the table by Dead Count and look for classes your code allocates You can also filter the table to help you narrow the search For details see Filtering Data in the Instances View on page 141 Often identifying a class or classes that are being allocated more than you expect is enough for you to know where to look in your program to change this behavior If not you can use JProbe to help you find the problem code 6 To get a snapshot of the data that JProbe has recorded click Take a Snapshot 3 H in the Runtime Summary view or click Data Recording amp and select
35. profiling data collecting data on methods 222 understanding 222 pruning subtrees 231 subtrees from the Call Graph 319 public method profiling 270 Q querying a set of instances 170 Quest Support requesting modules 67 R recording controls Coverage 372 data Memory 102 recursive nodes cumulative time definition 233 redo and undo buttons Performance 323 redoing navigate actions in the Call Graph 323 reducing objects over allocation 119 references 168 referrers 168 Instance Detail View 168 reference tree 168 registering node locked licenses 466 remote sessions running 459 starting 460 removing snapshots 449 triggers 429 renaming a settings file 63 categories 437 folders 437 reports Allocations 206 Allocations Detail 206 Call Traces 207 Coverage 399 Coverage Session 400 creating 440 Dominating Instances 205 Heap Graph 208 Instance Detail 207 Instances 204 Memory 203 Memory Snapshot Difference 208 Merged Allocation Points 208 Method Detail 347 Methods 346 Performance 343 Performance Snapshot Difference 347 Pinned Instances 205 Runtime Summary Memory 204 Performance 344 setting options Memory 204 Performance 344 Snapshot Browser 401 Summary 345 requesting node locked licenses 464 results communicating 439 root See loitering objects running Coverage analysis 359 Java application 72 Java EE application 73 Memory analysis 107 remote sessions 459 scripts 55 Runtime Instances
36. sessions on the same application Each concurrent license permits a developer to run a JProbe Analysis Engine and a JProbe Console Two people cannot connect JProbe Consoles to the same instance of a JProbe analysis session a separate instance of the application is run under JProbe for each concurrent user If no license is installed JProbe queries the local subnet for a valid license server on startup and adds it automatically if found Concurrent licenses are managed automatically If you attempt to use JProbe while all licenses are in use you are notified You need to wait for an available license If this occurs frequently you may want to contact Quest Software to increase your number of licenses The following procedures assume that you have completed these tasks Task Help on Task 1 Purchased JProbe for the operating To order contact sales quest com system that your application runs on and asked for concurrent licensing 470 JProbe User Guide Task 2 Downloaded the Quest Software License Server from the Web page specified in the confirmation email that you received and specified the host name and OS of the network server where it will run Help on Task To download the License Server follow the instructions in the confirmation email the License Details Web page and the License Server download Web page Your license file is emailed to you 3 Installed and started the License Server on
37. the group of encapsulated children of A is called by the name A and the group of encapsulated children of B is called by the name B and so on In the Call Graph these types of nodes are represented by an icon with an asterisk When the Collect Encapsulated Grouped option is selected the CCT that the Performance tool creates has a similar structure to the structure shown in the Grouped Encapsulation CCT diagram Grouped Encapsulation CCT Comparing the Grouped Encapsulation CCT to the Simple CCT diagram on page 223 the nodes for X2 and x3 are combined and their times and call counts are added together This type of encapsulation preserves the structure of the CCT and shows the boundary methods but without details The following diagram shows how the Call Graph represents this CCT 228 JProbe User Guide Grouped Encapsulation Call Graph Encapsulated None Finally with the Collect Encapsulated option set to None the encapsulated nodes disappear completely from the CCT and their times are added into their caller as displayed in the following example main 6 1 Encapsulation using None option This is the fastest type of encapsulation but there is no indication of boundary methods which leads to the situation that you might mistakenly think that a node called another node when really there are several nodes in between the calling node and the child node For this reason it is important
38. view appears It contains the following tabs e Memory Pools Tab displays memory usage in graphic and table formats e GC Data Tab displays garbage collection data in graphic and table formats e Threads Tab displays information about the application s threads Toolbar buttons provide access to the snapshot and garbage monitoring controls The system messages area bottom pane displays any errors and warnings that occur including those that have occurred before this Console connected to the JProbe Analysis Engine Exploring Performance Analysis Views 275 Exploring the Runtime Summary View welcome 9 Runtime Summary i Dp BF z 7 AYATA ru Pools and GC Interval One Minute Classes Loaded 2007 Classes Unloaded 0 Memory Pools GC Data Threads Memory Pools Memory Pools HEden space survivor space 4 tenured Gen Available 00 02 05 00 02 15 00 02 25 00 02 35 00 02 45 00 02 55 Memory Pool Distribution Used Mb Committed Mb Available Mb Maximum Mb l Total 13 0 18 8 5 8 167 3 Heap Subtotal 19 5 2 33 66 7 Eden Space I 0 0 0 9 0 9 42 Survivor Space I 0 0 0 1 0 0 os Tenured Gen I 18 4 2 24 62 0 Non Heap Subtotal 111 13 6 25 100 7 Code Cache I 1 0 1 0 0 0 33 6 Perm Gen f 10 1 12 6 25 67 1 Session Connected Finishing current recording and taking snapshot Snapshot C Quest_Software JProbe_8 3 snapshot_raw_1 jpp successful
39. 0 5 lt init gt String 100 0 3 initializeData InputStream 0 0 4 readWord 30 0 10 readLine 0 0 1 v Source Data You can add the source code to your report For each method the source code is displayed along with line numbers and whether or not the line was missed This report is useful to have with you if you want to talk to a developer about a particular method You may also find it useful for analyzing Coverage results away from the computer Reporting Coverage Analysis Results 405 Snapshot Browser Report Line information for method setDirections String Lines Missed Total Lines 100 0 4 Line Missed Source 107 tue direction 108 109 tue for int 110 tue director it 12 tue s new String numDirections rections x Report Filters By default the report includes data on all code in all packages in the snapshot You can choose to exclude some code from the report using report filters Adding report filters is the same as adding application filters or data collection filters To create a list of report filters 1 Click an empty cell and do one of the following e To select a package or class click the cell browse button navigate to the package or class and click Open e To select a method click the cell browse button and navigate to the class containing the method When you select the class you can expand its list of methods Locate and double click the desired method e To s
40. 100 0 M 100 0 aT 100 0 Adventure Ea Logic 23 6 ME 34 0 M 100 0 RoomList 20 0 E 26 1 E 100 0 Room gt 0 0 0 0 100 0 Bathroom 0 0 0 0 100 0 Show Only Methods with Missed Lines v gt 0 v Missed Lines Method demos coverage adventure Adventure lt init gt demos coverage adventure Adventure lt init gt java lang String No 0 0 0 9 demos coverage adventure Adventure cleanup Yes 100 0 M 6 6 demos coverage adventure Adventure initializeRoomList No 25 0 E 2 8 demos coverage adventure Adventure main java lang String No 50 0 M 7 14 demos coverage adventure Adventure play No 50 0 a 1 2 demos coverage adventure Adventure printUsage Yes 100 0 as 2 2 Catch blocks filtered You can filter the methods displayed in the lower table by using the Show Only Methods with Missed fields You can filter by the percentage of missed lines of code or by the percentage of missed conditions only methods that exceed the percentage are displayed For example if you choose Lines gt 50 the only methods displayed are the ones where over half of the lines of code in the method were missed by the test case A Closer Look at Data in the Application Tree The default view of the data shows a selection of percentage missed values for the application packages and classes listed in the Application Tree The following table describes the values in the tree The columns mar
41. 133 Performance 275 Memory Runtime Summary view interacting with 135 understanding the 132 Memory Source view 196 interacting with 197 opening the 197 Merged Allocation Points report 208 Merged Allocation Points view 120 181 interacting with 182 opening the 181 understanding the 181 Method Detail report 347 Method Detail view 325 calling context sites 330 customizing the 333 edges from parents area 328 edges to children 330 graph area 328 navigating the 332 opening the 326 relationship to Methods view 326 summary information 327 understanding the 326 method detail Coverage Snapshot Browser 381 Method Level filter 241 methods collecting data on 222 encapsulating data on 250 in Call Graph and Call Tree 249 signatures hiding 455 Methods List 305 Methods report 346 Methods view 297 Call Graph tab 298 Call Tree tab 302 customizing the Call Graph 323 editing the data set 318 Methods List 305 navigating the graph 320 metrics data collection filters and 241 threaded programs and 241 misses in Coverage Source view 385 in Snapshot Browser 378 modifying a test suite 356 multiple filters in Heap Difference View 200 N Native filter 241 navigating the Method Detail view 332 navigators child 247 parent 247 node locked license overview 464 registering 466 requesting 464 nodes hiding in the Call Graph 320 non switch conditions 386 O objects loitering 108 obsolete collection references 109 listeners 109
42. 279 Performance Summary view 292 Pinned Instances view 157 interpreting Heap Data 105 JProbe analysis results 414 Memory Pools 103 investigating Coverage results 363 deadlocks 264 performance bottlenecks 248 isolating subtrees in the Call Graph 319 J JAR file 43 Java application application arguments setting 33 class path setting 34 JVM setting 36 main class setting 32 running 72 working directory setting 32 Java EE configuring an application 24 running under JProbe 73 supported application servers 31 Java executable specifying 36 Java file locations setting 455 Java heap root set 108 Java options specifying 37 Java SE configuring an application 27 specifying an application 31 JBoss 86 87 JDBC component instrumentation for Performance 267 JProbe Analysis Engine 417 JProbe community 21 JProbe Connection Manager 417 JProbe Console background 417 Configurations Navigator 419 connecting to application started without JProbe 96 Java application running with JProbe 72 Java EE application running with JProbe 73 local application started without JProbe 96 remote application started without JProbe 97 interface 419 Navigator area 419 no connection 95 Snapshots Navigator 420 View area 419 JProbe options setting 47 JVM specifying for a Java application 36 L Leak Doctor view 190 interacting with 195 opening the 190 understanding the 192 license options 463 requesting 465 licensing concurrent 469 en
43. 338 0 v lt The values in the table represent the differences between the snapshots A negative value shows a reduction in time spent or objects created which translates to a performance improvement Understanding the Performance Difference View When you selected snapshots to compare you selected snapshots that represented similar session settings and were taken at approximately the same time in the session Also make sure that you have made the same edits such as prunes and isolates in both snapshots For more information Editing the Data Set on page 318 340 JProbe User Guide The Filter Methods fields enables you to filter data based on the package or class of a selected method By filtering methods you view only the methods that you are interested in investigating Caution Large percentage changes in methods with short method times may not be meaningful as even normal fluctuation in times may be seen as a large percentage increase or decrease The difference values are the result of a subtraction the compared snapshot minus the baseline snapshot A negative value represents a performance gain that is it means less time was spent and or fewer objects were created in the compared snapshot A positive value means more time was spent or objects created A value of zero means no change If you have percentages turned on you can see how great a change occurred The percentages are calculated using the
44. 390 Where Were Connections Closed SQL Statements SQL Method cats CPU Cml Elapcmi Y SELECT num addr FROM location ORDER BY num java sql Statement executeQuery java lang Stri 1 127 update location set num addr where num java sql PreparedStatement executeUpdate 2 x 30 insert into location values java sql PreparedStatement executeUpdate 2 28 What Next To investigate a connection or statement double click its table row If the snapshot does not contain any JDBC data the Summary view presents only the information relevant to Methods The following illustration presents an example of a Summary view when no JDBC data is recorded and only the Elapsed Time data is available A welcome Runtime Summary IE Summary 7 ia Performance Snapshot Taken Sep 28 2009 2 53 58 PM Timers Elapsed Recorded Level Full Encapsulation Elapsed Execution Time of All Threads Combined A 15s Recorded Time 5s Performance Hotspots Methods to Investigate Name Method Time of A demos performance diners Philosopher run 99 2 a a a java lang Thread sleep long 0o71 java io PrintStream printin java lang String on java awt Graphics drawImage java awt Image int int java awt image ImageObserver 0 0 java lang Thread start 0 0 Methods That Match Session Filters to Investigate Name Method Time of A demos perform
45. 416 o o java util Hashtable Entry 1 24 825 19 800 0 0 Instances Report This report contains the content of the table in the Instances view Before generating this report sort the table by the metric that you are interested in Reporting Memory Analysis Results 205 Dominating Instances Report Instances by Heap Count Recorded Recorded Reape oe Keep Alive Dead Dead ae Count oxen count Memory Size Bytes Count Memory Bytes Bytes Bytes demos memory sim Sim Result 5 000 80 000 5 000 80 000 80 000 0 0 java lang String 3 72 2 801 67 224 186 832 10 060 241 440 char 0 0 2 666 169 424 169 424 0 0 java lang Class 0 0 1 169 489 944 774 832 0 0 java util HashMap Entry 0 0 607 14 568 22 640 81 1 944 java lang Object 0 0 339 40 840 123 544 0 0 java util Hashtable Entry 1 24 315 7 560 21 904 Q 0 java lang Integer 0 0 307 4 912 4 912 0 0 java lang Short 0 0 256 4 096 4 096 Q 0 java lang String 0 0 211 8 344 17 968 0 0 sun java2d loops Blit 0 0 199 7 960 7 960 Q 0 sun java2d loops ScaledBlit 0 0 163 6 520 6 520 0 0 java util concurrent locks ReentrantLock NonfairSync 0 0 145 3 480 3 672 o 0 java util concurrent ConcurrentHashMap Segment 0 0 144 4 608 12 904 0 0 Tamas E ine eke 0 0 144 2 680 8 264 0 0 Dominating Instances Report This report contains the content of the table in the Dominating Instances view Before generating this report sort the table by the metric that you are interested in Pinned Instanc
46. 7 Opening the Pinned Instances VieW ssessssssssessessscsessessecsessecsessesssensseseusensseeeuessesaeeseeansansansensateateateaneaneeneenes Understanding the Pinned Instances ViCW ssscsssesesssesesesnessssseestsseestestssessnssnssessessesenseessateneeateateanenteaneens Interacting with the Pinned Instances View Exploring the Allocations Vie W ar nan a aaa aeaaeai aa ana rahaaa ihien ERER Opening the Allocations View measina a A A A a a A Understanding the Allocations VioW ssesssssecsssssssesssssesssssessesseessecsessesseesessesseeneestesseneeseeseeateseeseeneeseesteatesseneeess Interacting with the Allocations View sessessessessessesesssssessesseeseesteetessessesseeseeneeseeneensesteneeateatesteaeesteatenteaeesenss Exploring the Allocations Detail Vi W c cecsssscessesestessecesseseseeseeesseeeeseeseeasseeaesesseeassneassesseeasseeaseneaneasaneaseneaeeeseneases Opening the Allocations Detail View t i seisan aaar er EA EAE EEEE E AEDA Understanding the Allocations Detail View Interacting with the Allocations Detail VieW ssssssesssssessessessessesstestsstesessesesenseateatentestsateaneateateneenseateaneateaeers Exploring the Instance Detail View 2isicciasi dive anesia chin ae adnan Opening the Instance Detail Vie t cavansescecsieeinenneii cane aiea anes natn vinta aia Understanding the Instance Detail View 0 0 csessessesesseesssssssesstestssnssnsstesnsseesnsenseateaseneestest
47. 84 welcome pages hiding 454 working directory specifying for a Java application 32 Z zero values in the Snapshot Difference view 339 486 JProbe User Guide
48. 89 H Heap Data 134 filtering 105 interpreting 105 Memory 105 Heap Difference View understanding data 199 heap dump HPROF importing 129 IBM JVM importing 127 importing 127 Sun JVM 129 Heap Graph report 208 Heap Graph view 185 interacting with 187 opening the 185 understanding the 186 hiding nodes in the Call Graph 320 subgraphs in the Call Graph 321 hits in Coverage Session 375 hotspots critical path 244 definition 220 finding 243 how JProbe works 416 HTML file exporting to 442 443 IBM WebSphere AS 83 85 WebSphere AS Community Ed 82 WebSphere Portal Server 84 identifying deadlocks 260 importing heap dumps 127 478 JProbe User Guide IBM HPROF files 129 IBM PHD files 127 Instance Detail report 207 Instance Detail view 166 instance analysis 185 Memory interacting with 171 opening the 166 referrers and references 185 understanding the 167 instance list 167 Instances report 204 Instances view 138 filtering data 141 opening the 138 understanding the 139 integrating configurations 55 interacting with 336 388 Allocations Detail view 164 Allocations view 161 Call Traces view 178 Coverage Session 376 Dominating Instances view 152 Heap Graph view 187 Instance Detail view 171 Leak Doctor view 195 Memory Difference view 200 Memory Runtime Summary view 135 Memory Source view 197 Merged Allocation Points view 182 Performance Difference view 340 Performance Methods view 305 Performance Runtime Summary view
49. Action Description Recording Starts recording data In the Initial Recording at JVM Start section you can select the following options e Data Recording Off e Data Recording On Note When the trigger action is set to Recording the selected recording parameter is also displayed in the triggers table 372 JProbe User Guide Trigger Action Description Snapshot Takes a snapshot The snapshot file name takes the form snapshot_ lt n gt _raw jpc Recording Controls You can use the data recording controls to start or stop recording data for a session by setting them in the Automation tab The default setting is Data Recording On During a session you can start or stop recording data using the toolbar buttons If you want to use the toolbar button to control when data recording starts set the Initial Recording at JVM Start option to Data Recording Off For details see Controlling a Session with Runtime Toolbar Buttons on page 374 18 Exploring Coverage Analysis Views This chapter provides detailed information about the Coverage views used to present coverage data This chapter contains the following sections Exploring the Coverage Session Exploring the Snapshot Browser Exploring the Coverage Source VOW iscsi sicsscicsi ise vatiiobesenisnsndisisndniesavitins 384 Setting Preferences for Coverage VIOWS ssssssssssscsesessesssssssstecntssnesseentesnesseesstsneentessees 390 Learning More About Coverage
50. Apache Felix configuration file config properties The config properties file is located in the GLASSFISH_HOME glassfish osgi Vfelix conf directory You must manually add these filters into the config properties file before integrating GlassFish v3 with JProbe For example you can update the property by using the following command org osgi framework bootdelegation javax transaction javax transaction com jprobe Note An error message is displayed when no com jprobe filter strings exist in the Apache Felix configuration file Starting Your Application with JProbe Attached 89 Runtime Notes for Application Servers When you integrate your configuration JProbe creates a startup script resaves your settings file and also makes a copy of the GlassFish domain xmil file called lt configuration_startup_script_name gt _domain xml In lt configuration_startup_script_name gt _domain xml calls to java are replaced with calls to jplauncher Caution You should make a backup copy of the original domain xm file If you encounter a problem you can manually restore and load your copy of the domain xmI file When you run the JProbe generated startup script you can specify parameters that you usually pass to the start domain command of asadmin bat Do not however specify the domain name on the command line because it is set in the configuration If you do specify a domain name the application server will not run corre
51. COMMGUIRIIONS iiit aaa aa 59 What Can Do if My Application Server Is Not Supported ssssisssssssssissrsrrssssrsrrerssssrsrrerssssrsnne 64 24 JProbe User Guide Configuration Overview The following sections describe the process of creating a JProbe configuration for your application and attaching JProbe to obtain data for analysis Before you can do this you or your system administrator must have completed the following tasks as outlined in the JProbe Installation Guide e Installed JProbe on your workstation and specified a valid license e For a network installation also installed JProbe on the server where your application runs and specified a valid license e Ifthe JVM that you use to run your application or application server is not supported installed a supported JVM on the computer where your application runs Configuring a Java EE Application A basic setup takes only a few minutes though it may take longer if your application is complex or if you want to read more about the available options Caution Not all application servers are supported on all platforms For more information see the JProbe Installation Guide Before you begin you need to have the following information e If your application server has a startup script the path to that script e If there is no startup script the path to the server s home and or bin directory e The names of packages used in your application e The type of analys
52. Columns to Fit The columns are resized Export the table to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Exploring Performance Analysis Views 297 Exploring the Methods View Exploring the Methods View The Methods view is a powerful tool for viewing detailed information about the focused method selected in the Summary view and its calling relationship with other methods in your application Its features are discussed in the following sections e Opening the Methods View e Understanding the Methods View e Interacting with the Methods View e Customizing the Call Graph Opening the Methods View The Methods view can be reached from the Performance Summary view To open the Methods view e Inthe Summary view double click the method you want display The Methods view opens with the selected method as the focus method The view has two ways to visualize the calling relationships a call graph and a call tree as well as a summary view of the data in a method list which contains a list of your methods and threads if included with data for all the metrics Understanding the Methods View The Methods view consists of two sections the upper pane contains the Call Graph and Call Tree tabs The lower pa
53. Configuration and Integrating on page 55 Defining a JProbe Configuration You can think of a configuration as containing everything necessary to run your application under JProbe A JProbe configuration contains the following information For Java EE applications the application server you use and details about how it is set up in your environment For Java SE applications the main class working directory class path application arguments and JVM you use The code that you want to focus on in the analysis 30 JProbe User Guide e The type of JProbe analysis Memory Performance or Coverage that you want to run on your application and specific options for that analysis type e Advanced JProbe options optional When naming a configuration you may want to include words that make it easy to distinguish among your configurations You could specify the application name the application server Java EE or the JVM Java SE used and the type of analysis For example configuration to analyze the memory usage of the JProbe Leak Example demo running on WebLogic 10 could be named GamePack_WebLogic10_Memory Choose the specific Java EE application server on which your application runs by selecting its name in the Select the Configuration Type tree B New Configuration Wizard Configuration Type Define a New Configuration Provide a name for this configuration Configuration Name GamePack_Memory_WebLogic10 Select
54. Data Recording Off A new memory snapshot is added to the Snapshots Navigator You can open this snapshot in the Instances view by right clicking the snapshot and clicking Open Snapshot Select Dead Count or Dead Memory from the Investigate by list then and select from the table the class you found that is being over allocated Examining Garbage Collection Data Two views are especially useful for finding a problem the Call Traces view and the Merged Allocation Points view Both views show the call traces for garbage collected objects but from different perspectives The Call Traces view shows the data in a traditional Call Graph or Call Tree much like the Call Graph in JProbe Performance except with object allocation data instead of times You may find this view more helpful in understanding the overall allocation behavior The Merged Allocations Point view shows the data starting from the allocating method This view shows the merged allocation traces for a set of objects which may be more helpful in quickly identifying the cause of a problem Using the Merged Allocation Points View The Merged Allocation Points view works almost the same way with garbage collection traces as it does with other recorded data see Exploring the Merged Allocation Points Running a Memory Analysis 121 Tuning Garbage Collection and Finding Over allocations View on page 181 except that all the objects have been garbage collected so you can
55. Diner 2 cr Dirar 4 F 2 2 2 a demos performance dners Diner 1 lt int Diner 4 0 4 4 4 0 demon performance drar Diner 3 crd Diner 4 3 2 2 2 o demos performance diners Oiner checkJOsVersont 4 v 4 4 4 0 demos performance divers Diner mari Stringi J 4 2 029 2 2008 2 o demos performance dners Diner 3 componertResizediComponertEvert 2 4 4 4 o demon performance draers Chopstik Cet gt Phdosopher anal rt 45 P 2 2 2 o demes performance diners Philosophers ank gt Piloseeher 45 9 4 4 0 0 demas performarke draers PhiosagharP mei Loof equest a 2 2 4 o demes performance diners Philosopher soel 40 1 4 4 0 0 demas perfamaxe drers PhiosoghaP mei tartRequest s 13 2 2 4 o demes performance dners Diner 1 actorPerformedi Actor vert 16 0 4 4 4 o demas perfamarxe drers Phiosogher startin 0 3 a 2 4 o demes performance dners Phiosogherg runt 206 0 1 4 4 o demos performance drers Phiosoghar lt ink gt ChepStick ChapSick PhiosaghaPana s5 4 0 o demes performance diners PhiosopherP anei parti Graphics 201 1493 4 1 0 o demos performance ders Phicsopher andi gee hioscohar sand hapshds 4 7 a J Gemes performance diners Phiosagher partiGr aphies 1 005 im 3 0 o demos performance drers Phiosoghar Panel amp saPhiosophersi a 19 19 3 2 o Gemes performance diners Prdosogher anei crk gt Diner 1 3 32 3 0 demos performance diners Diner lt int gt a 200 1s 200
56. Do These Pin option from the list The Pinned Instances view appears Understanding the Pinned Instances View The Pinned Instances view allows you to view the instances that are being held in the heap by a specific set of instances referred to as dominators and as a consequence cannot be garbage collected Dominators offer a different way of viewing heap data based on reachability analysis If every path from roots in the heap must pass through one instance for example a Widget before getting to another for example a String the former is said to dominate the latter the Widget dominates the String During snapshot processing JProbe analyzes the paths in the heap and identifies which instances are reachable only through other instances It builds a tree structure called a dominator tree which starts with the roots of the heap and includes all instances in the heap Using the Pinned Instances view you can analyze what s pinning an instance by navigating towards the leaves of the tree The Pinned Instances view consists of three panes 156 JProbe User Guide Welcome IS Instances Hi Pinned Instances ELAI Dominating Instances by Instance Type Name Recorded Count Heap Count Recorded Memory Heap Memory Keep Alive Size Y 2 038 849 560 1 380 392 java lang Class Pinned Memory by Pinning Type Categories ava Base Libraries Hav
57. Dominator The upper pane Pinned Instances by Instance Type contains a table which shows details about the instance set selected in the Instances view Each row represents the set of pinned instances that you selected as input to the view This is the starting set of instances from which the data in the rest of the view is derived These instances are grouped by class similar to the Instances view table The middle pane Pinned Memory by Dominating Type shows the categories of instances responsible for pinning memory in the starting set The pane contains a chart which illustrates the percentage of dominating instances by Keep Alive Size in each of the categories that are listed in the legend to the right of the chart Hovering over a segment of the chart displays a popup showing the memory kept alive by that category Selecting a bar in the chart redraws the Dominating Instances view using instances from the selected category as the starting set This allows the investigation of why instances are live to continue towards the roots that pin the Java heap The lower pane Pinned Instances by Dominating Type contains a table which shows the breakdown of classes responsible for pinning the instances listed in the upper pane The values displayed in the Recorded Count Heap Count Recorded Memory Heap Memory and Keep Alive Size columns in the lower pane represent a subset of the total shown in the corresponding columns in
58. Elapsed times can also be affected by the system resources available to your program such as processor speed other running applications available memory and so on Within bounds the more resources the faster your program will run Obviously this means that your program s performance will differ from one computer to another and across platforms CPU Time Use CPU time when you want to separate the performance of your code from how your runtime environment handles your program CPU time does not include program pauses such as I O and is not affected by system resources with the exception of the timing resolution provided by the operating system itself CPU Time is often more expensive to collect than Elapsed Time so your program may run noticeably slower under JProbe when collecting CPU Time as compared to Elapsed Time Note Some operating systems do not offer CPU timing In these cases JProbe automatically uses elapsed time CPU time measures the number of processor cycles spent executing code in all threads The resolution of the CPU timers is determined by the operating system and is often much coarser than the resolution of elapsed time When a resolution is comparatively coarse it is entirely possible that a very small method and certainly single lines of code will execute within a single tick of the timer As JProbe can only measure at the unit of 258 JProbe User Guide resolution provided by your operating syst
59. If you are renaming a settings file so that you can keep your old settings file and then change your analysis options the preferred way do this is to copy your configuration For more information see Copying a Configuration on page 59 To save settings to a settings file 1 Select the configuration in the Configurations Navigator 2 Inthe lt Settings_File_Name gt tab click Save As or Right click the configuration name and select Save As The Save JProbe Settings File dialog box appears 64 JProbe User Guide 3 Type in a directory and file name for the new settings file and click Save 4 To update your startup script to call the new settings file click Integrate 5 Follow the prompts to save the files for this configuration What Can I Do if My Application Server Is Not Supported If your application server is not currently in the list of supported application servers you may still be able to run JProbe with it Caution Unsupported application servers are untested JProbe may encounter unforeseen issues while collecting data on your application You have three options for configuring an unsupported application server The following table lists your options and states the advantages and disadvantages of each approach Option Advantage Disadvantage Creating a Configuration for an You can create a configuration right now using Unsupported Application Server the Java Application option The disadvantage is
60. Import IBM Text File e Import IBM PHD File 3 From the Select IBM Text PHD File dialog box browse to the file that you want to import 4 Click Open The dump is converted to a Memory snapshot 128 JProbe User Guide 5 Select and view the Memory snapshot as usual How to generate a heap dump from an IBM JVM Some IBM JVMs automatically generate a heap dump when an out of memory exception is thrown You can also generate a heap dump in situations where you are using an IBM JVM and are unable to take a snapshot You need to set the environment variables to allow heap dumps on request and to specify the format of the heap dump with the exception of the out of memory situation described below You can generate a heap dump by sending a signal to the process CTRL on UNIX and CTRL Break on Windows IBM_HEAP_DUMP and IBM_HEAPDUMP must be set before taking a heap dump The following environment variables allow the JVM to generate a list of all the objects in the Java heap IBM Dump Task Action Add Java environment On your system set the environment variables variables IBM_HEAP_DUMP true IBM_HEAPDUMP true IBM_HEAPDUMP_OUTOFMEMORY true IBM_JAVACORE_OUTOFMEMORY true IBM_JAVA_HEAPDUMP_TEXT true If you set the environment variable IBM_JAVA_HEAPDUMP_TEXT true the JVM creates a text heap dump If you set this environment variable to false the JVM creates a PHD heap dump The default setting is false Text dum
61. Match Session Filters to Investigate Name Method Time of A demos performance diners Philosopher run 97 demos performance diners Philosopher paint java awt Graphics 0 0 demos performance diners PhilosopherPanel drawPhilosophers 0 0 demos performance diners PhilosopherPanel paint java awt Graphics 0 0 demos performance diners Philosopher 1 run 0 0 What Next To investigate a method in one of the tables double click its table row To investigate any method enter its name here classname methodname v Investigate Understanding the Summary View The Summary view contains the following tabs e Methods Tab displays information that helps you identify potential performance hotspots and focus on problem methods e JDBC Tab displays information related to JDBC connections and statement methods Important This tab is displayed only if JDBC data is recorded in the snapshot Toolbar buttons provide access to the execution log and to the report creation dialog Methods Tab The Methods tab is pre populated with information that helps you identify potential performance hotspots and focus on problem methods From this tab you can navigate to subsequent views which provide detailed information about the selected methods of interest 286 JProbe User Guide The Application section upper part of the view presents global performance data about the session that was run to obtain the snapshot
62. Ne ee Ge R E cate 368 Filters for Coverage ce ute ose ats lt at NG ES CL ss 369 Triggers Tor Coverage i sesiterencstrusr eth ca dhasaanih WM Stina iain ia 371 Exploring Coverage Analysis VieWS sssssssesesusuuuresnsuuuuurunnununnunennunununnnonunununnnonunununnnonnnunnnnnonnnnnnnnonannnnenennann 373 Exploring the Coverage Session edecctttisa dices eeiaidhish stad nian eeainienatenuincvel evunaundauie 374 Controlling a Session with Runtime Toolbar Buttons cessessesssesteeseseesecseeseesesstessteseeseesneeseeneeeseeeeneesnees 374 Understanding the Data in the Coverage Session ssssessssssssessecsecsessesssesessnesssesenessnsaeeatssteansatsensateatesteaseess 375 Interacting with the Coverage S SSION csstssecesssessesseesessesseesecstecseesneeseecntesecatesneeaerstesseeseseesiecneeseeneesees 376 Exploring the Snapshot Browsets s ias a e Ar a Eae AE pa ar AE aE a A EE e EEEE eaaa Eia EARS 378 Opening the Snapshot Browser seseesseessecsesssscssecnessecseesseesecseesseesessteseeseeeseesneeseeeateaseeateesseneeatereteaseesteneees 378 Understanding the Data in the Snapshot BrowSer ss sssssessessessessessessessesseesesesessesasenessnsaessnssnsarsensseseeeaneeteaseess 378 Interacting with the Snapshot BrowSel ssesscsecseseesesseseesesseesneeseeseesneesecneesneesteeneesneesesseesneeteesieeneesneenesees 382 Exploring the Coverage Source View s cs cssccssecscssecseecsasesscassocsstcanrsvscaseneaecasesnsdasn
63. Opcode Opcode Stack Arguments Description Mnemonic 153 ifeq value true if equal zero 154 ifne value true if not equal zero 155 iflt value true if less than zero 156 ifge value true if greater than or equal to zero 157 ifgt value true if greater than zero 158 ifle value true if less than or equal to zero 159 if icmpeq valuel value2 true if int equal 160 if_icmpne valuel value2 true if int not equal 161 if_icmplt valuel value2 true if int less than 162 if_icmpge valuel value2 true if int greater than or equal 163 if_icmpgt valuel value2 true if int greater than 164 if_icmple valuel value2 true if int less than or equal 165 if_acmpeq valuel value2 true if reference equal 166 if_acmpne_ valuel value2 true if reference not equal 198 ifnull value true if reference is null 199 ifnonnull value true if reference is non null Interacting with the Source View You can perform the following tasks Exploring Coverage Analysis Views 389 Exploring the Coverage Source View Task Action Response Navigate to a block of Click a stripe in the The block of missed code missed code Coverage Bar is displayed Navigate to the code Click the down arrow on The next block of code represented by the next stripe in the Coverage Bar the Coverage Bar or right click and choose Next Stripe missed lines or missed conditions is displayed Navigate to the code represented by the previous stripe in the Coverage Bar
64. Recording Level feature and the Heap Data tab which are not available you can interact with the session in the same way as in any other Memory analysis session such as requesting garbage collections and taking snapshots 100 JProbe User Guide 3 Strategies for Memory Analysis Memory is an analysis tool that helps you to improve memory usage and reduce unnecessary garbage collection in your Java applications As you run the session you use the Runtime Summary view to see the overall memory behavior of the program and try to spot classes that may be leaking or are over allocated After you capture the data in a memory snapshot you use the heap investigation tools starting with the Instances view to identify leaking instances and find where they were allocated or to find the source of over allocated objects This chapter provides recommendations for how to use the JProbe Memory analysis tool in your development environment Tip We recommend analyzing memory usage before measuring performance This chapter contains the following sections RO O Dele sv eias aawk ae aaaih ian hadi tadv i eas ne ee ee ee 102 Pantie SUMAN aeaaaee a Eaa etaan aana dawned 102 102 JProbe User Guide Recording Data After you have started your application and connected JProbe to it you can set the level of data that you want to record e Record Allocations For All Instances e Record Allocations and Stack Traces For All Instances e Record Allocat
65. Results 0 0 esessssesssessseseeseseesstesneesneestesntesneeateesnesteeneesnees 393 374 JProbe User Guide Exploring the Coverage Session The Coverage Session provides a real time view of test coverage for your application as your session is running The Coverage Session remains available when the session ends you can toggle between it and other views This section contains the following information e Controlling a Session with Runtime Toolbar Buttons e Understanding the Data in the Coverage Session e Interacting with the Coverage Session Controlling a Session with Runtime Toolbar Buttons Data recording is set when you create a settings file on the Analysis Type gt Automation tab The default setting is Data Recording On which means that JProbe starts to record data as soon as your application s JVM starts If you choose Data Recording Off you can use the toolbar buttons in the Coverage Session to control when JProbe collects data For example you may want to get past all the startup activities in your application before beginning to collect data In this case you click the Start Recording toolbar button just before you begin your test case For more information see Recording Controls on page 372 In addition to starting and finishing data collection for the test case you can take a snapshot of data at any time during the session The following toolbar buttons are available during the session Icon Acti
66. Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 9 Optional Specify the JProbe advanced options For more information see Specifying Advanced JProbe Options on page 46 In this screen you have the following options e Continue the configuration process by reviewing the settings Click Next and continue on step 10 e Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 e Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 10 Optional Review the settings in the configuration summary screen For more information see Reviewing the Configuration Summary on page 49 The configuration is now complete In this screen you have the following options e Edit any of the configuration settings if necessary Click Edit beside the setting that needs further updates The wizard opens the screen corresponding to the selected setting then guides you through the appropriate configuration steps Caution You may need to define again some settings in the subsequent wizard screens for example filters and elements e Save the configuration settings
67. Signatures c essssesseseesesseseesesseeseesesssesneeseesnsesueesesneeeteesieeneeseeneestenenaee 455 Specifying Source File LOCatiONS versessersen0exwevescnea nisensenasaeannvawimiiasnnira ERAON 455 14 JProbe User Guide Configuring Memory POools cisersseccotei nicer eeseie E A Nailin whiadbneanihnatinas 456 Running Remote Sessions ssssssssssssssssessssssesessesseesssseesesssesssssesnseneeseeeesesssesesaseaesaeeaeeaseaeeaseaeenesseeetnaneennanees 459 Starting and Viewing a Remote Session sscescesssseesseessetesseesssessstesseeseseesstessatseaeesstessaessueessessanessaeeeaeesanessneesay 460 WProbe License Options icsacsesstescsfchcueds fei cites te rete cet cas i 463 Node Locked Licensing sic catrasdtaraiaeiiura aceial aa aianii artic E 464 Requesting Node Locked Licenses 5 csscisssssssssssesessasessssasssstesassanssassansvecescssanssassaetaasssansssssaanssensncesesannsensantsiesines 464 Registering Node Locked Licenses s ssessessessssessessecsesetesescsessesesseeseesseneeseentesseseeaseneeseenseseeseeasestenteneneesees 466 Per S at LICENSING sissies Anwediravi Aine dina acon O ENA EEEE A REEE 468 Concurrent LICEMSing faces es RE R ANN AAA R NN 469 Using Concurrent LOEn SEE aa ie ea ea aT 469 EE eA Te nae AEEA EAEE EA aussi ins E AA 472 Introduction to This Guide This chapter provides information about what is contained in the JProbe User Guide It also provides information about t
68. Unloaded Memory Pools GC Data Heap Data Memory Pools Memory Pools 6 eden space survivor space 4 tenured Gen Available 00 01 45 00 01 55 00 02 05 00 02 15 00 02 25 00 02 35 Memory Pool Distribution Used Mb Committed Mb Available Mb Maximum Mb Total 7 0 18 5 11 5 167 3 Heap Subtotal 0 5 5 2 4 6 66 7 Eden Space I 01 0 9 0 8 4 2 Survivor Space 0 0 01 0 1 o5 Tenured Gen 0 4 4 2 37 62 0 Non Heap Subtotal 6 5 13 3 6 9 100 7 Code Cache I 0 7 0 8 0 0 33 6 Perm Gen al 5 7 12 6 6 9 67 1 Session Connected Finishing current recording Snapshot C Quest_Software JProbe_8 3 snapshot_raw_1 jph has been added to the snapshot navigator as snapshot Finished recording snapshot C Quest_Software JPrabe_8 3 snapshot_raw_1 jph successfully written Memory Pools Tab The Memory Pools chart shows the history of the used and available values of the pools The Available part is the sum of the displayed pools Available memory While you are running an analysis session the summary graph displays the size of the heap memory pools by default Heap pools contain things normally found in the Java heap like Java objects Non heap pools normally contain things the JVM uses internally such as the compiled representation of method code or class information To change from viewing heap to viewing non heap data or vice versa use the Configure
69. aaa aaa 108 Tuning Garbage Collection and Finding Over allocations c escsssssssssssssssssessnssssesssssserseees 117 108 JProbe User Guide Finding Memory Leaks This section discusses memory leaks in terms of loitering objects and walks you through a basic Memory analysis looking for loitering objects This section includes the following topics e Understanding Loitering Objects on page 108 e Setting the Recording Level on page 110 e Performing a Basic Memory Analysis on page 112 Understanding Loitering Objects The key to fixing and avoiding loitering objects is good reference management Loitering objects are objects in the heap that are reachable through references from the root set but have no future value to the program They are caused by a live object inadvertently or unnecessarily retaining a reference to the loiterer for some period of time To allow the garbage collector to remove a loitering object and possibly further instances that it references you must remove all references to the loiterer Memory leak in C C handled by JVM s GC Memory leak in Java Some objects are reachable from the root set while others are not A root is an initial entry point to the Java heap the root set includes all entry points to the heap for your program There are several ways that objects can be part of the root set including but not limited to being one of the following Running a Me
70. allocate even if they do not cause garbage collection problems can hurt performance Understanding Garbage Collection Issues Garbage collection will always have some impact on your program and the type of garbage collector affects the impact Some garbage collectors can cause long pauses that are noticeable and annoying in programs with graphical user interfaces but acceptable in server applications Others minimize the pause time but this can lower the throughput of server applications Modern JVMs use ergonomics to try to adapt garbage collector behavior automatically to work best with your program They may allow you to specify a goal such as low pause times or high throughput and will pick an appropriate garbage collection strategy Often this automatic behavior is good enough but you can use the Runtime Summary view to help you decide if you need to do further tuning See Monitoring GC Overhead with the Runtime Summary View on page 118 for details One key to achieving good GC behavior that is common to all JVMs is correct sizing of the Java heap Many collectors are generational garbage collectors because they divide the heap into pools for newly allocated objects and older objects that use different garbage collection strategies Getting the relative sizes of the generations right is also important for good overall garbage collector performance 118 JProbe User Guide Monitoring GC Overhead with the Ru
71. analysis session The Runtime Summary helps you to identify loitering objects excessive garbage collections and over allocation of objects and also acts as a Strategies for Memory Analysis 103 Runtime Summary progress bar It is the only view available to you until you take a snapshot during or at the end of the session When you attach to a session the Runtime Summary appears automatically It has three tabs e The Memory Pools Tab displays memory usage data e The GC Data Tab displays garbage collection data e The Heap Data Tab allows you to examine the classes and instance counts in the Java heap of your application as well as recorded and garbage collected instances tracked thus far Toolbar buttons provide access to the snapshot and garbage monitoring controls For a detailed description of this view see Exploring the Memory Runtime Summary View on page 132 Memory Pools Tab The Memory Pools tab in the Runtime Summary view displays detailed data and summary data about memory usage The chart displays the available size of the Java heap allocated memory and the memory in use by your program over time You can adjust the interval for data display that is updates to heap non heap and garbage collection data as well as control whether the chart displays heap or non heap data A table below the chart provides data for each memory pool Interpreting the Memory Pools Graph The Memory Pools graph helps you to visu
72. and about the snapshot itself The following table lists the type of information displayed in the Application section Parameter Information Provided Notes lt ApplicationName gt Name of the application N A being analyzed Snapshot Taken Date and time when the N A snapshot was taken Recorded Level Performance recording N A level used when capturing the snapshot Recorded Time Actual recording time N A clock time Timers Timers available Elapsed N A or Elapsed CPU Elapsed Execution Time Sum of all threads N A of All Threads Combined A elapsed cumulative times minus the sum of all threads elapsed method times CPU Execution Time of All Threads Combined B Sum of all threads CPU cumulative times minus the sum of all threads CPU method times Available only when collecting CPU data CPU Utilization Rating CPU utilization rating value and icon Good OK Bad or Severe X Available only when collecting CPU data Exploring Performance Analysis Views 287 Exploring the Summary View The CPU Utilization Rating is defined as the CPU Execution Time of all Threads Combined divided by the Recorded Time Four CPU Utilization Rating ranges are defined as follows Range Rating 0 5 lt CPU Utilization Rating Good 0 25 lt CPU UR lt 0 5 OK 0 1 lt CPU UR lt 0 25 Bad CPU UR lt 0 1 Severe The Perfo
73. are important to the analysis Note In the JProbe Console you can select filters via the Set Recording Level dialog The filters table in the dialog contains a default filter to track all allocations made by your application Running a session with this default filter helps you to identify problem classes in your application If you know which classes are causing your problems you can set up JProbe to collect more data on them To collect stack traces as well as allocation data define filters for the problem classes and set the action to Include Caution Collecting detailed information on stack traces and garbage collections requires a lot of resources Set these filters on problem classes only not your entire application Filters are defined in the same way for all the analysis tools however the filter actions differ among the analysis tools This section summarizes the filter actions for a Memory Learning More About Memory Options 125 Analysis Options for Memory analysis To learn how to create filters see Refining Data Collection Filters Filters for a Memory analysis can be set on a package or on a class Filter Action GCs Description JProbe collects data about object allocations stack traces and garbage collected instances for the specified package or class In the JProbe Console data collected with this action is displayed in the Merged Allocations Points view and the Call Trace view when investigating
74. are saved to the same directory do not separate these files When you run the JProbe generated startup script you can specify parameters that you usually pass to the WebSphere startup script Do not however specify the server name on the command line because it is set in the configuration If you do specify a server name it is ignored To choose a different server copy the current configuration change the server name and integrate Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Usage lt configuration_startup_script gt options where options are clean Restores WebSphere to a pre JProbe state after an abnormal termination e nowait e quiet e logfile lt filename gt 86 JProbe User Guide e replacelog e trace e script lt script filename gt background e timeout lt seconds gt e statusport lt portnumber gt e recovery e help Before starting the server the startup script updates the server configuration settings to use jpjava executable instead of java executable After starting the server the script restores the server configuration settings If the server does not start success
75. buttons to move backward and forward through operations you have already performed or you can click the down arrow beside either button and go directly to any of the other operations that you have performed as shown below Similar to the Instances view you can sort the tables by any of the columns The text field above the table allows you to filter the classes that are listed in the table See Filtering Data in the Instances View on page 141 for details Filtering Data in the Dominating Instances View In the Dominating Instances view the filters set under the Set Preferred Dominators icon Y are used to determine which dominators are worth investigating during the memory analysis In some cases you may only be interested in the instances that your code is pinning If filters are set to include your code the dominators for each instance are searched for instances that match that filter If an instance has a dominator that matches the filter that dominator s class is considered a preferred dominator and presented in favour of the 152 JProbe User Guide instance s immediate dominator in the table in the lower pane as well as in the middle pane s chart If the filter was not matched the instance s immediate dominator is used instead Interacting with the Dominating Instances View The following table summarizes the tasks that you can perform Task Action Response See the instances pinned
76. by the selected dominators Select a class and click Instances Detail View OR Select a class right click and select the Show This Dominator s Pinned Instances option from the list Note This action is available only if the selected classes have instances that are live in the heap The Instance Detail view opens with all the instances for the selected class listed in the upper pane Focus on instances pinned by your code Click Set Preferred Dominators Y The Set Filters dialog box opens The classes in the table reflect the filtering expression s you have entered For more information see Filtering Data in the Dominating Instances View on page 151 Show immediate dominators Click Show All Immediate Dominators The lower pane is updated to display all immediate dominator instances Exploring Memory Analysis Views 153 Exploring the Dominating Instances View Show preferred dominators Click Show Preferred Dominators The lower pane is updated to display only the dominating types defined as preferred through the use of filters If a preferred dominator is not found its immediate dominator is shown See the instances that dominate a set of instances Select a class right click and select the Why Are These Live option from the list The Dominating Instances view opens showing what instances are dominating pinning the selected
77. called choose Method Invocation Based Trigger e To execute the trigger based on elapsed time choose Time Based Trigger Refining Data Collection and Presentation 431 Setting Triggers 3 Choose when and for how long the advanced trigger is active e To execute the trigger once after a specified number method calls or minutes choose Perform the trigger action only once and specify the number of methods call minutes e To execute the trigger after a specified number of method calls or minutes and every time the method is called thereafter choose Delay performing the trigger action and specify the number of methods call minutes e To execute the trigger at specified intervals for a set duration choose Customize when the trigger action is performed and specify values for the start the interval and the end e To remove an advanced control from a trigger choose Default control 4 Click OK An icon in the row button for the trigger indicates that an advanced control is set Adding a Trigger in the Body of a Method Normally triggers can only act on entry or exit to a method If you need a trigger in the body of a method this provides you with greater control on executing the method you can add an empty method to use as a trigger point and place a call to it anywhere in your method under any conditions you can express in Java code Then when using JProbe set a trigger on the added method For example if you have a method like
78. can perform from the Method List context menu Task Add a method to the Call Graph Action select a node in the Method List right click and select Fit Method Into Graph Response The method is added as a node to the Call Graph Any other previously hidden nodes that are required to link the new node to the existing graph are also added Move to the Method Detail view with a focus method Open Method Detail View The Method Detail view opens with the selected method as the focus method JProbe User Guide Task View the list of threads represented by this node Action select a node in the Method List right click and select View Thread Names Response Opens the Threads dialog box which contains a list of thread names represented by this grouped node View the list of methods represented by this node View Method Names Opens the Methods dialog box which contains a list of method names represented by this grouped node Drill down to source code Show Source If you have the source code the Source view opens with the source code displayed Drill down to line Show Line Data The Source view data associated with the selected method opens Disabled if line data is not available for the selected method Set filter based on Set Filter Based On The Method List is the method s e Method Package updated to display only package or cl
79. chart Therefore this feature is unavailable In the JProbe preferences you can choose whether or not headers are included in the data For more information see Changing How JProbe Exports to a CSV File on page 455 To export a runtime view chart 1 Open the Memory or Performance Runtime view 2 To export the chart data right click in the chart and choose Export to CSV 3 Specify a fully qualified file name Tip You may want to add the csv file extension to the file name for clarity 4 Choose Heap Non Heap or Both from the Pool Type drop down list 5 Click OK Saving a Chart as an Image You can save a chart as a JPEG or PNG file and add that image to your reports Set up the chart the way you want it before saving it as an image Note The Coverage views do not include charts Therefore this feature is unavailable To save a chart as an image 1 Open the view containing the chart 2 Set up the view 444 JProbe User Guide 3 Right click in the chart and choose Save Chart As gt lt image_type gt where image type can be JPEG or PNG 4 Specify a location and file name 5 Click Save Viewing the Execution Log When running a session you can access the execution log from the Runtime view toolbar The same information is also saved to every snapshot taken during the session The following table summarizes the information in the execution log Tab Name Description General Contains information abo
80. configurations and the package level application filters that you specified for each of them 2 Choose packages and click OK JProbe User Guide 424 The chosen packages are listed in the Data Collection Filters table The filter action is set to the default action for the chosen analysis type The new filters are enabled by default Changing Filter Execution Order Filter order is important When two entries refer to the same code the entry lower in the table is used A red line appears under filters when there is a conflict Task Move a filter Action Click the button at the left of the filter s row and drag the filter to the new location Response The filter is inserted where you drop it Insert an empty row between two filters Right click the button at the left of the filter s row and choose Insert The filter is inserted above the chosen filter row Removing Filters You can remove filters in one of two ways If you think you may use the filter again you can temporarily disable the filter The filter will not be used in this session Otherwise you can delete the filter Caution Order is important when using filters Be careful when disabling filters that you do not inadvertently remove something that you want to analyze For more information see Changing Filter Execution Order on page 424 Task Action Response Disable a filter Clear the check box beside the f
81. configuring my environment Contact Quest Support for assistance When you do be prepared to describe your environment The customer service representative may ask you to run a JProbe session that includes some debugging options to describe your settings more fully My program called a method in a derived class but don t see the method in the Call Graph This issue arises when you set an encapsulated filter on the derived class but use a method or line level filter on the base class What happens is that when you call a method in the derived class through a reference of the base type JProbe attributes the performance data for the derived method to the calling method The easiest fix is to ensure that both classes use the same type of filter Alternatively you can set a line level filter on the calling method to see how much time was spent calling the derived method Consider the following program snippet and imagine that Base has a method level detail filter while Derived uses an encapsulated filter class Base public static void main String args Base base new Derived base method public void method System out println In Base method class Derived extends Base public void method System out println In Derived method In this example all performance data for Derived method is attributed to main and Derived method does not appear in the Call Graph or Method List If you cha
82. contained in the snapshot 2 From the Investigate by list select Recorded Count Recorded Memory Dead Count or Dead Memory 3 Select from the list a class that has data in the Recorded or Dead columns and click Call Traces EK The Call Traces view opens Note The No Recorded Instances dialog box appears when the selected class contains no recorded instances that is when either the JVM did not provide traces for the instances or trace collection was not enabled when the instances were allocated A Welcome amp Runtime Summary Ba instances EX Call Traces T Chr to Show Cumae co w OBE zE 42 hel y i e Cunulabve Court Y Cumulative Memory Method Court Method Memory Source Merged Traces Tree Traces Graph Nace Cumdative Count Cundative Memory Y Method Count Method Mear yeveteng Ihreadrunt 5 000 w 000 o o My Code 5 000 40 000 5 00 40 000 Investigating Recorded Irstarnes Understanding the Call Traces View The Call Traces view provides you with several ways of viewing the same method data It is a good starting point for investigating which methods allocated what instances 176 JProbe User Guide The upper pane contains two tabs e Merged Traces Tree is a traditional program call tree that shows methods that are involved in the allocation that is it lists the call tree of the program for the call paths that allocated instances that are still live e Tra
83. coverage adventure Room setDescripti No 0 0 demos coverage adventure Room setDirection No 0 0 demos coverage adventure Room setName int No 0 0 demos coverage adventure Room setNumDirec No 0 0 demos coverage adventure Room showDescrip Yes 100 0 demos coverage adventure Room showDirectio Yes 100 0 Session Connected Finishing current recording and taking snapshot Snapshot C Quest_Software JProbe_8 3 snapshot_raw_S jpc successfully written Snapshot C Quest_Software JProbe_6 3 snapshot_raw_5 jpc has been added to the snapshot navigator as snapshot Session Disconnected You can filter the methods displayed in the lower table using the Show Only Methods with Hit Lines field Note The Show Only Methods with Hit Lines field is enabled only if you have selected Line or Line with Condition as the coverage level option when defining the Coverage analysis configuration This option filters the results by percentage of hit lines only methods with a percentage of hits below the chosen percentage are displayed For example if you JProbe User Guide choose lt 70 the methods with 0 to 69 hit lines are displayed while the methods with 70 to 100 coverage are hidden Tip You can change the color used for hit lines For more information see Customizing the Colors Used to Display Results on page 392 Interacting with the Coverage Session You can perform the following tasks
84. displayed It contains the port number that JProbe is using for this session Note In some cases the JProbe Execution Console window does not open and the application server log files show the JProbe Port number The application server log file is located in a directory specific to each application server 74 JProbe User Guide gt JProbe Execution Console C Quest_Software JProbe_8 Seal ce EET x sing JPROBE_HOME C Quest_Sof tw nbs Seite ising JPROBE_JPL jp_input C Do c JProbe Execution A C Quest_Software JProbe_8 3 demos gamepack p jpinput4929071311251358774 jp1 sing JAVA HOME C Quest_Sof tware JProbe_8 3 bin jre ising GERONIMO_HOME C Quest_Sof tware JProbe_8 3 demos game pack geronimo tomc t sing GERONIMO_TMPDIR var temp sing JRE_HOME Cc Nguese_ Sof tware JProbe_8 3 bin jre Probe version 8 3 6 copyright 2089 Quest Software Inc ALL RIGHTS RESERVED Probe is a registered trademark of Quest Software Inc overage analysis started WProbe Connection Manager is available for Console connections on port 52991 jooting Geronimo Kernel Cin Java 1 6 _12 tarting Geronimo Application Server v2 1 4 boseneccoocHEEcHEHEEEEHEEEEouEeccsoeeee 188 26s Startup complete Listening on Ports 1050 0 CORBA Naming Service 6 RMI Naming Derby Connector OpenEJB ORB Adapter OpenEJB Daemon OpenEJB ORB Adapter Tomcat Connector AJP AJP Tomcat Connector HITP BIO HTTP Tomcat Connector HTTPS BIO
85. do in any class named C in any subpackage of P except P directly matches P SubPkg C doGet but does not match P C doNotMatch Pre Pret All methods in all classes starting with Pre in any package as well as any class in a subpackage whose name starts with Pre matches both Pkg Prepare m and Pkg Pretend C m Configuring JProbe to Analyze Your Application 43 Specifying Your Code If you type this Canonical formis And filters are applied to this s y s y All methods in classes whose name ends with y and either starts with s or belongs to a subpackage that starts with s matches both java rmi server RMISocketFactory com quest say and java security cert xX509CRLEntry foo bar foo bar All methods in the class bar in package foo This is likely not what you wanted See the next example foo bar foo bar All methods in all classes in foo bar or any of its subpackages foo bar String t e no change Methods whose names start with t and end in ein foo bar String matches both toUpperCase and toLowerCase Elements Table You can specify elements of your application and have JProbe create filters based on the packages and classes that they contain For this purpose elements are defined as WAR files JAR files EAR files and any directory containing exploded code Tip For Java Applications when defining your classpath you
86. e One thread Thread A calls method e It then synchronizes on lock_1 but may be preempted at that point e The preemption allows another thread Thread B to execute e Thread B calls method2 e It then acquires lock_2 and moves on to acquire lock_1 but cannot because Thread A has lock_1 e Thread B is now blocked waiting for lock_1 to become available e Thread A can now resume and tries to acquire lock_2 It cannot because Thread B has acquired the lock already e Thread A and Thread B are blocked The program deadlocks Of course most deadlocks are not so obvious simply from reading the source code especially if you have a large multithreaded program Performance can identify deadlocks and point out where in the source code these deadlocks originate so that you can fix them Hangs That Are Not Deadlocks In some cases your program may hang even though it does not contain a deadlock situation The difference between a deadlock and other kinds of hangs is that in a deadlock situation the threads can never move forward whereas in a hang the threads will not move forward The most common example of a hang is when a thread has called a wait method but no thread ever calls the notify method to wake it up Consider a program that has a producer thread and a consumer thread while done Object item producelItem synchronized queue_ queue_ add item queue_ notify Running a Performance Analy
87. e Source Data e Report Filters Application Summary The Application Summary provides an overview of your application For each package it shows the total number of classes methods and lines of code and the percentage of each that were missed by your test suite This summary is useful for communicating Coverage results to senior management Packages Name Classes Missed TotalClasses Methods Missed Total Methods Lines Missed Total Lines Total 33 3 6 45 2 31 58 1 210 demos coverage adventure 33 3 6 45 2 3 58 1 210 Reporting Coverage Analysis Results 403 Snapshot Browser Report Package Summary The Package Summary breaks down the results in the Application Summary by package For each class in a package you get the total number of methods and lines of code in the package and the percentage of missed methods and lines You can use this summary to get an overview of test case coverage This summary is also useful as a management tool you can assign classes with a high percentage of missed code to members of your team for further investigation Packages Name Classes Missed TotalClasses Methods Missed Total Methods Lines Missed Total Lines Total 33 3 6 45 2 3n 58 1 210 demos coverage adventure 33 3 6 45 2 31 58 1 210 Per class summary for package demos coverage adventure Name Methods Missed Total Methods Lines Missed Total Lines Total 45 2 a 58 1 210 Adventure 42 9 7 63 9 61 RoomList 20 0 5 41 4 2 L
88. family size and style Use the Font Chooser dialog to change the font family size and style of text in the graph To change fonts 1 In the toolbar click g to open the toolbar menu and select Set Font The Font Chooser dialog box opens Font Chooser Name Style Dialog v Regular Sample Text Select a new font from the Name list Select Regular Bold or Italic from the Style list Select a font size from the Size list Your choices are displayed in the sample text below the list boxes Click OK Interacting with the Heap Graph The following table summarizes the tasks that you can perform Task Action Response Change the color Select an option from the The graph changes to the scheme of the graph Color to Show list other scheme Enlarge or reduce Move the button on the Scale The graph becomes larger the graph slider or smaller JProbe User Guide Task Move to a section of the graph that is not in the viewing area Action Use the scroll bars to move to the hidden section or Click Graph Panner and move the rectangle to the desired area Response The graph viewing area shows the selected area View details of the selected instance Click Instance Detail E The Instance Detail view opens Investigate the selected node Click Leak Doctor GY The Leak Doctor view opens View the immediate dominator of a selected instance Se
89. filtering data 105 Runtime Summary report Memory 204 Performance 344 Runtime Summary view Memory 102 132 Performance 274 S saving chart as an image 443 configurations 49 50 55 snapshots 448 snapshots on exit 450 Index 483 script running 55 session history viewing 445 set of instances queryinga 170 Set Recording Level Memory 110 Performance 236 setting advanced trigger controls 427 Coverage options 360 current snapshot 448 Memory report options 204 Performance analysis options 266 Performance report options 344 trigger execution time 429 settings file creating a configuration from a 61 integrating 55 renaming 63 saving 49 50 55 showing nodes in the Call Graph 320 parent subgraphs in the Call Graph 322 previous view of the Call Graph 323 subgraphs in the Call Graph 321 top children in the Call Graph 321 top parents in the Call Graph 322 snapshot 418 baseline for heap comparison 198 baseline saving 239 basename 48 compared 198 IBM Dump 127 129 IBM HPROF file 129 IBM PHD file 127 overview 418 Snapshot Browser 378 Snapshot Browser report 401 Application Summary 402 Class Summary 403 Package Summary 403 Source Data 404 484 JProbe User Guide snapshots loading 449 managing 447 saving 448 saving on exit 450 setting current 448 transfers managing 451 unloading 449 Snapshots Navigator adding snapshots to list 449 overview 420 source file locations specifying 455 SpringSource dm Server 94 stack
90. following formula Percentage Compared Baseline Baseline 100 If the resulting value is gt 1000 or lt 1000 the change is reported as huge If a method was not called in the baseline snapshot but is called in the compared snapshot the percentage is replaced with a plus sign to indicate a new method Interacting with the Performance Difference View You interact with this view in the same way as with the Method List in the Methods view You can show metric columns sort the table by a metric column use the Find feature and use the Filter feature For information on how to use the Performance Difference view see Measuring Performance Improvements on page 258 The following table lists the tasks that you can perform Task Action Response Filter methods by Select a method right click and The table refreshes to package select Set Filter Based On gt show only the methods in Method Package that package or Type or select a filter in the Filter Methods field Exploring Performance Analysis Views 341 Exploring the Performance Difference View Task Filter methods by class Action Select a method right click and select Set Filter Based On gt Method Class or Type or select a filter in the Filter Methods field Response The table refreshes to list only the methods in that class Show hide table columns Right click and select Show Hide Columns Select the
91. following table lists the subgraph related tasks that you can perform Child Navigator Task Hide a subgraph Action Click the child navigator are shown default action if all top children Response The node remains but all visible methods called directly or indirectly by the method are hidden Restore a hidden subgraph Click the child navigator action default after a Hide Subgraph Restores the visible methods in the previously hidden subgraph Navigating Calling Methods Use the parent navigator to hide and show the methods that called a selected method The term parent navigator refers to the triangle displaying a plus or minus sign at the left side of a node The following table lists the parent related tasks you can perform Parent Navigator Task Show only the subgraph for the selected method and hide all other nodes Action default action if other subgraphs are shown Click the parent navigator Response The method is selected and all direct children and indirect descendants calls to methods are displayed All other nodes are hidden Restore nodes removed by showing only a subgraph Subgraph action Click the parent navigator default after a Show Only The parent methods and their trees are displayed Exploring Performance Analysis Views 323 Exploring the Methods View Going Back to Previous Views of the Graph
92. from the Investigate By list 2 Select a class that has instances 3 Click Merged Allocation Points Ba The Merged Allocation Points view opens displaying with the merged stack traces for the selected class If the message Empty Dataset no matches appears either the JVM did not provide traces for the instances or trace collection was not enabled when the instances were allocated Understanding the Merged Allocation Points View The Merged Allocation Points view is a good starting point for investigating where instances originate Each method that allocates instances of the selected class becomes 182 JProbe User Guide the root of a tree The tree represents the merged stack trace for the method A merged stack trace includes all methods that called that particular allocating method Beside each allocating method you can see the number of instances allocated by the method in the Count column O82 Not Avaliable 818 Not Avaliable Cumulative Count Y 882 O Norm OtjectEverander new Object java lang Object o2 818 Investigating Recorded instances Note You may see the same method twice in the list Duplicate allocating methods occur when two lines of code in the method each allocate instances When you expand a stack trace you can see the series of method calls that led to the allocating method being invoked The arrows beside the method names show the direction of the calls toward the allocating m
93. information code and action JProbe proposes default data collection filters based on how you described the code you are interested in see the My Code screen of the Create Edit Settings wizard You can modify the data collection filters to increase or reduce the amount or complexity of data collected Caution Filter order is important For more information see Changing Filter Execution Order on page 424 The procedure for adding filters is the same for all the analysis tools however the filter actions differ among the tools Note You can add data collection filters during the configuration wizard on the Specify JProbe Analysis screen or when editing the configuration via the Analysis Type tab To add a filter 1 Inthe Runtime Summary view click the Set Recording Level icon on the toolbar then select Record Data as Specified in the Table Below option to activate the Data Collection Filters table 2 Click an empty cell in the Data Collection Filters table and do one of the following e To select a package or class click the cell browse button navigate to the package or class and click Open Refining Data Collection and Presentation 423 Refining Data Collection Filters e To select a method click the cell browse button and navigate to the class containing the method When you select the class you can expand its list of methods Locate and double click the desired method e To select code from an element su
94. information see Editing a Configuration on page 60 e Integrate your application to run with JProbe Click Integrate For more information see Integrating Your Application on page 51 e Run your application with JProbe Click Run For more information see Running Your Script on page 55 e Run your application with JProbe and connect the JProbe Console Click Run amp Connect For more information see Running Your Script and Connecting the Console on page 58 Configuring JProbe to Analyze Your Application 51 Integrating Your Application Integrating Your Application When you select Integrate JProbe creates a startup script JProbe also verifies that the configuration s settings file is still available and re saves the settings For some application servers a copy of the application server s configuration file may be created and edited Saving the JProbe Startup Script To save the JProbe generated startup script 1 Inthe Integration Wizard dialog box on the Integrate tab review the integration status This information is saved to the Configurations log file which is located in the JProbe temporary files directory the default on Windows systems is Documents and Settings lt user_name gt jprobe lt JProbe_version gt lt computer_name gt Tip You can right click in the text box to select and copy the text to another location Integration Wizard Integrate Integrating GamePack
95. instance set See the instances pinned by the selected dominator Select a class right click and select the Show This Dominator s Pinned Instances option from the list The Instance Detail view opens and displays the instances pinned by the selected dominating type Calculate Actual Keep Alive Size The sign in the Keep Alive Size column indicates an estimated value for this metric To calculate the actual size right click the row and select Calculate Actual Keep Alive Size from the list The Keep Alive Size column is updated with the actual Keep Alive Size value Save the chart as a PNG or JPEG file Right click in the chart and select Save Chart As then select the graphic file type PNG or JPEG in which you want to save the chart In the Export Image dialog box specify the export parameters and a location to save the file The file is saved to the location you indicate 154 JProbe User Guide Show hide table Right click the table and select The selected columns are columns Show Hide Columns In the displayed The rest are Show Hide Columns dialog box select the columns to show and click OK hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and sele
96. lines of code and missed conditions if collected If you collected data at the Line with Condition level and the snapshot contains data on conditions you can choose a condition and see details about it in the Condition Details table Coverage Bar Methods Adventure main java lang String welcome Runtime Browser fj Snapshot Browser Adventure java v Show Data For v Lines v Conditiops Line Missed Missed Conditions Source 106 107 109 110 111 112 113 114 115 116 117 118 119 120 Yes Adventure adventure null Av if args length gt 1 System out print1n Adventure class getName Too many arguments provided printUsage System exit 1 switch args length game where directions are er from the console windot adventure new Adventure Conditional data Opcode if_icmple True Yes False A Memory Performance a Coverage Catch blocks filtered A Closer Look at the Data in the Source Code By default the source code is color coded to highlight missed lines of code and if available missed conditions 385 e Missed lines including condition statements with all outcomes missed are red e Condition statements with some missed outcomes are gold e Code that was hit by your test case is black e Code for which
97. list in the Snapshots Navigator 450 JProbe User Guide To unload all snapshots 1 Choose File gt Unload All Snapshots 2 Ifsome snapshots have not been saved you are prompted to save them e To save some unsaved snapshots select the check boxes beside the snapshots and click OK Save the snapshots as usual e To save all unsaved snapshots click Select All and click OK Save the snapshots as usual e To delete all unsaved snapshots click Clear All and click No e To leave the dialog box without unloading snapshots click Cancel All snapshots are unloaded and the Snapshots Navigator is empty Saving Snapshots on Exit Snapshots are temporary you need to save them before exiting or they are deleted By default you are prompted to save temporary snapshots when you exit JProbe Tip You can disable the prompt to save temporary snapshots For more information see Controlling How JProbe Operates on page 454 To save multiple snapshots on exit 1 Choose File gt Exit to close JProbe If there are unsaved snapshots the Save Snapshots dialog box opens listing all the temporary snapshots 2 Choose the snapshots you want to save Save Snapshots Unsaved snapshots exist Select any you wish to save LeakExample_Memory snapshot_S LeakExample_Memory snapshot_6 zy Z Leake xample_Memory snapshot_ B Polynomial snapshot 7 OK cancel Help _ selectat Cl
98. maximum heap size with a specific value such as Xms512m Xmx512m For example the compiling process for JSPs takes up a significant amount of memory If for example your session is failing at a load of only five pages per second this is probably a memory allocation problem Running through the JSPs at a slow pace allows the application server to cache them and afterwards you can increase the load on the system without difficulty In the Runtime Summary view the memory reported in the Memory Pools Summary chart does not match the total line in the Runtime Data tab The value reported in the Memory Pools Summary chart represents an approximation of the size of the heap based on summary data received from the JVM about memory used and memory freed Because it is an approximation it may not exactly match the total in the Data tab Troubleshooting Memory Analysis Results 213 Memory Troubleshooting Tips So for the purposes of your investigation use the chart and values in the Runtime Summary as a starting point but for serious investigation take a snapshot and use the Instances view to analyze your data The Merged Allocation Points or Call Traces view states Empty data set No matches If the Merged Allocation Points view indicates Empty data set No matches this may be because the JVM did not provide traces for any of the objects of the class Another reason may be that you did not set the runtime session settings t
99. name The button is disabled when no threads are selected or a category is selected You can detect deadlocks on demand by clicking the Detect Deadlock button If a deadlock is detected the threads list displays the threads involved in the deadlock If no deadlock is detected a message is displayed to that effect Exploring Performance Analysis Views 279 Exploring the Runtime Summary View Interacting with the Performance Runtime Summary View The following table summarizes the tasks that you can perform Task Change the data recording level or add a filter Action Click Set Recording Level i to open the Set Recording Level dialog box then make the desired changes Response JProbe collects the selected level of data or applies the filter specified in the table The current data set is saved to a snapshot A new snapshot is started unless Data Recording was set to Off Take a snapshot Click Take Snapshot w A snapshot of heap is taken It appears in the Snapshots Navigator Take an HPROF heap dump Click Take an HPROF Heapdump Be An HPROF dump is taken converted to a JProbe snapshot and added to the Snapshots Navigator You can then analyze it using the JProbe Memory analysis tool Request a heap garbage collection Click Request Garbage Collection T A garbage collection operation is performed and the results are displayed in the GC Data tab View the execution lo
100. object and class initialization methods This section describes the features of the Source view e Opening the Source View e Interacting with the Source View Opening the Source View If you have the source code for a method you can open the Source view from a context menu or from the toolbar To open the Source view 1 Select the method that is a row in the Methods view or Method Detail view 2 Right click and select Show Source or Click Show Source oe or Right click and select Show Line Data if the selected method has line data collected The Source view opens with the code for the selected method displayed If you collected your data using Line Level filters the Source view also displays line level detail in additional table columns 336 JProbe User Guide Welcome GF Runtime Summary R3 Summary Methods Pl Method Detail amp Philosopherjava Methods demos performance diners Phukasogtrer cunt une Cals Eao Method Dap Cmi Source i 5 1 1 Thread thisThread Thr E de A 67 5 o o String name thisThread getHame s C s 0 o im eatSpeed 8 o a v o while omer se thisThread nm a 3 x sleepVaiue int Math random panel controller eleepTine 7 l 3 s Systen out printin name waits for Chopstich ch id 7 try 73 i 2 1 121 Thread sleep sleepValue i 1 it own 1 thisThread ie 0 0 0 break a 4 508 3 908 synchronized eno n l 0
101. often a trigger action is performed The two types of triggers are Method Invocation Based Trigger and Time Based Trigger The options for each are the same with some differences in the results For examples of how to use these options see Refining When a Trigger Executes on page 429 The following table summarizes the options available for selection and describes what actions are performed with each type of trigger 428 JProbe User Guide Option Method Invocation Based Time Based Default Perform the trigger action Perform the trigger action every Control every time the method is time the method is executed executed Perform the trigger action Perform the trigger action on the n th execution of the Perform the trigger action the first time the method is executed n only once method where n is choosen minutes after program start where from the drop down list n is choose n from the drop down list Delay Start performing the trigger Start performing the trigger action performing action on the n th execution the first time the method is the trigger of the method and then executed n minutes after action perform the trigger action program start and then perform the every time the method is trigger action every time the executed where n is choose method is executed where n is n from the drop down list choose n from the drop down list Customize Start performing the trigger Start performing
102. only if JDBC data is recorded in the snapshot The following table lists the operations that you can perform in the JDBC tab Task Investigate a connection Action Double click a row in the Connections table OR Select a row in the Connections table right click and select Investigate Response The Methods view opens with the selected graph node associated with the row as the focus method Investigate closed connections Select a row in the Connections table and click the Where Were Connections Closed button The Methods view opens with the Connection close method as the focus node Investigate a statement Double click a row in the SQL Statements table The Methods view opens with the selected graph node associated with the row as the focus method Exploring Performance Analysis Views 295 Exploring the Summary View Task Investigate where a statement was executed Action Select a row in the SQL Statements table right click and select Where was this Executed Response The Methods view opens with the java sql Statement execute method as the focus node Investigate where a statement was Select a row in the SOL Statements table right click and The Methods view opens with the collected select Where was this java sql Statement Collected execute method as the focus node Tune an SQL Select a row in the SOL The SQL Optimi
103. other trees remain in the graph though the number of calls attributed to the nodes decreases by the number of calls associated with the pruned tree in this case by one each 320 JProbe User Guide Original Call Graph After Pruning Subtree a _ Note Line level data is not recalculated While the Call Graph displays the newly recalculated method level data the Source view continues to show the original line data Navigating the Graph by Showing and Hiding Nodes You can change the view of the graph without editing the data set JProbe keeps a history of changes to the view and also a history of edits to the data set For more information see the following topics e Hiding and Showing Nodes on page 320 e Navigating Subgraphs on page 321 e Navigating Calling Methods on page 322 e Going Back to Previous Views of the Graph on page 323 Hiding and Showing Nodes You can hide and show a node and you can show a single node with only its top parent and child methods Note Hiding and showing nodes does not change the data set To remove some data from the data set use isolate or prune For details see Editing the Data Set on page 318 Exploring Performance Analysis Views 321 Exploring the Methods View Ti D SI I The following table lists the node related tasks that you can perform Task Hide a method node Action Click the X on the
104. percentage of missed conditions Methods with Missed Conditions greater than the chosen percentage are shown The rest of the methods are hidden Exploring Coverage Analysis Views 383 Exploring the Snapshot Browser Task View the source code fora class or method Action Choose the class or method right click and choose View Source You may be prompted to locate the source code file Response The source code is displayed in a source view See Exploring the Coverage Source View on page 384 Sort a table by a metric Click the column heading for the metric The table is sorted To reverse the sort order click the heading again Show hide columns Right click and choose Show Hide Columns Choose the columns to show and click OK The chosen columns are displayed The rest are hidden Rearrange columns Drag a column by its column heading and drop it in a new location The column appears in its new location Resize column widths e To set the width drag the edge of the column heading left or right e To size a column to fit its data right click the column and choose Size lt column gt to Fit e To size all columns right click and choose Size All to Fit The affected columns are resized When sizing all columns to fit if the resized columns do not fill the table the width of all columns is increased to fill the empty space Export table da
105. pool that must be displayed in the chart and clear the rest of the check boxes Setting Preferences 457 Configuring Memory Pools 4 Click OK The chart refreshes to display the selected memory pools 458 JProbe User Guide Running Remote Sessions JProbe runs your program with the JVM you specify and listens for local or remote TCP IP connections from the JProbe Console JProbe uses a Connection Manager to act as a proxy for connections to both local and remote sessions The Connection Manager simplifies the procedure for connecting the Console to the Analysis Engine and enables you to attach and detach the Console to and from a running session using the toolbar buttons In a remote session there are two installations of JProbe one on the server and one on the workstation You set up and start the JProbe session on the server where your application runs then you connect to the session from a workstation For example with a Web application you configure a session to analyze Java Server Pages executing on a server but you can sit at your own workstation to run the client browser and interact with the session Note Usually you run a remote session to analyze Java EE applications However you use the same procedure to set up a standalone application that you want to monitor from a remote computer This chapter describes how to run remote JProbe sessions It contains the following section Starting and Viewing a Remote SeSSION
106. recording off and takes a snapshot To end the analysis click Detach From Session F on the JProbe Console toolbar The Instances view opens displaying the data in the snapshot taken before detaching from the running session You can now examine the data in various ways to identify loitering objects For a detailed description of the Instances view see Exploring the Instances View on page 138 Running a Memory Analysis 117 Tuning Garbage Collection and Finding Over allocations Tuning Garbage Collection and Finding Over allocations The discussion in Understanding Loitering Objects on page 108 explains how your program s runtime performance can be adversely affected by the garbage collector In addition to eliminating loitering objects with Memory and resolving performance bottlenecks in your code with Performance you can assess the behavior of the garbage collector with your program Tuning garbage collections is dependent on the JVM that you are using to run your program Although garbage collectors all perform the same basic function the way they operate varies widely among JVMs and even within a JVM JProbe can help with some general analysis and some tasks that are common but for complete tuning you will also need to consult your JVM vendor s documentation A closely related issue is excessive object allocation JVMs have many optimizations for allocating small inexpensive objects but objects that are expensive to
107. references stalled 110 stalled stack references 110 starting a remote session 460 strategies achieving program efficiency 218 subgraphs in the Call Graph navigating 321 subtrees isolating in the Call Graph 319 pruning from the Call Graph 319 Summary report 345 Summary view 283 support 21 support bundle generating 446 SupportLink requesting a license 465 switch conditions 386 T tables exporting data toa file 442 tabs Call Graph 298 Call Tree 302 GC Data Memory 134 Performance 276 Heap Data 134 Memory Pools Memory 133 Performance 275 Threads 276 technical support 21 temporary files location changing 454 test cases running Coverage 361 test suite creating 356 modifying 356 text conventions 20 threaded programs and performance metrics 241 threads in Call Graph and Call Tree 249 Threads tab 276 time measuring 257 Tomcat 79 80 traces 170 transferring snapshots 451 transitional references lingering 110 triggers adding 425 advanced controls 427 collecting data 256 collecting data using 256 Coverage 371 execution order changing 429 managing 425 Memory 125 Performance 271 removing 429 setting execution time 429 within the body of a method 431 troubleshooting Coverage analysis results 409 JProbe Console does not connect 95 Memory analysis results 211 Performance analysis results 349 U understanding the Allocations Detail view 163 Allocations view 160 Call Traces view 175 Dominating Instances vi
108. relevant to you Task Show hide table columns Right click and select Show Hide Columns Select the columns to show and click OK Response The selected columns are displayed The rest are hidden 334 JProbe User Guide Task Resize column widths Action To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit Response The columns are resized Sort the table by a metric Click the column heading for the metric The table is sorted To reverse the sort order click the heading again Export the table data to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers Exploring Performance Analysis Views 335 Exploring the Performance Source View Exploring the Performance Source View When you open a performance snapshot in the Source view you see your source code with line numbers If you collected data at line granularity for any methods in this source code you will also see columns for performance data by line Available metrics for line level data include Calls Elapsed Method and Elapsed Cml for the line Note Line data is not shown for
109. ssacsashencoastvanesassaussatesns canshatesedsarasisiaensaciatetanssntaiesined 353 No Line Dataline ua irene din eee OTEA A 353 Strategies for Coverage Analysis s sssssssssssssssssssssessseseesssseesesseenessseetsatesenaseeeesssenessseneeseeaeeneetenesateateaneennans 355 Assessing an Existing Test Suites asscrsevcswesnnranan neue adarnnianiian ae EEE EE 356 Table of Contents 11 Determining Overall Coverage sc iccnarcshieatieievacnsncavan tania etn diee a tiawbaie 356 Moditying a Test Suites aani anal aa aaah KEEN E aiid 356 Grating a New TeSt Suite a ate ces a te at ca Uae eee te ie a ace ld 356 Keeping Up with Developmennt s cscsicsarsstzsseascaacaessezsseisscaasessanssassntasesssatessssnsacessscnsassatsancassennsaeenteeaestsuasapeansanaenuanagc 357 Running a Coverage Analysis ssssssssessssssssssssssesssesesssesessnescenesseenesstesseaessnaseeseasseneeaseaeenseaeeneeeeneeatenteateeseans 359 Setting Up JProbe for a Basic Coverage Analysis cssesecseseseseseesesseeseestesseesteesesnsesusensesnseatesnsenteenseneeneennees 360 Running Test Cases with JPrObe a2 dics anata reveals nanan i a a E ENa 361 Investigating Your Result Ann Acai en A 363 Learning More About Coverage Options ssssssssssesssssssssesessseseessesessenessseeesessssnessseaesatesteaneetnaeeneatsateeteans 367 Analysis Options for Coverage ici cevesnicaccmmachiee Role iain Lacan eats Baar cain 368 COVETAGE OPtlOMS a see sesso ios es a ee ee i
110. startup script from the JProbe Console 1 Inthe JProbe Console select a configuration from the Configurations Navigator The lt Configuration_File_Name jpl gt tab appears in the View area 2 Click Run The Run lt Configuration_Name gt dialog box appears 56 JProbe User Guide Run Demo Diner Performance Analysis Path of script to start a java application or server with JProbe CiV Quest_Software Prabe_8 3 demos settings Diner_Performance_Analysis bat Script Parameters Working Directory CiViQuest_Software JProbe_6 kd V Close syscon window on exit Required Field Run amp Connect Cancel Help 3 Define the following run configuration settings e Select the JProbe script to start your application server The startup script field is pre populated with the latest startup script saved for the selected configuration Note Ifthe startup script file needs any environment settings you should set these environment settings before executing the jpconfig or the jpconsole application For example if the script file needs a JAVA_HOME value you should first set this value in your environment then run jpconfig or jpconsole and finally proceed to run the startup script file e Optional Type in any script parameters that should be used when launching the script e Select your working directory where the startup script resides e Optional Select the Close syscon window on ex
111. that match the provided name You can use the filter any time during a running session When you filter the data an indicator beside the filter displays the number of classes displayed out of the total number For details about filter syntax see Syntax for Instance Filters on page 142 Note The memory value in the Totals line of the Runtime Summary table may not match the memory usage reported in the Instances view because of the differences in how the memory is tracked for each For your investigation you should focus on the values in the Instances view For more information see Memory Troubleshooting Tips on page 212 106 JProbe User Guide Running a Memory Analysis The most common reason to analyze memory usage is to find memory leaks In Java a memory leak refers to objects that loiter in the Java heap because they are not available for garbage collection Another reason to analyze memory usage is to improve the efficiency with which the garbage collector runs This chapter describes some of the causes of loitering objects It also describes how to run an analysis session locate and investigate loitering objects and presents some techniques for solving memory problems The procedures are general but the screenshots and examples are taken from the LeakExample class program installed in the lt jprobe_home gt demos memory leakexample directory This chapter contains the following sections Fading Momay Leake ocsi
112. that you are familiar with your own program For example B did not call D though it might appear that this is the structure from the tree Isolating and Pruning Call Trees This section describes how isolating and pruning trees affects the data that you see in the Call Graph Strategies for Performance Analysis 229 Understanding Performance Data Isolating Subtrees In the Call Graph data about all calls to a particular method are combined into a single node Included in this method is data that you might not be interested in viewing For example consider the following scenario You are using a Java EE application server and begin to profile code Your code is called by its doGet method and your methods may call StringBuffer append You are interested in how much time was spent by your code in the append method The application server itself also calls this method and it is difficult to tell what time is the result of your own code or the use of the application server s The CCT allows you to remove the application server s use of the append method by using the Isolate command on the doGet method The Isolate command makes a new CCT rooted at the selected node To illustrate this command start with the following CCT maa pesa Gaz foi CCT for Isolate The corresponding Call Graph looks like the following CCT 230 JProbe User Guide Y X3 2 2 CCT for Isolate via Call Graph
113. that you may miss something and the configuration does not work The JProbe Community Website ttp jprobe inside quest com also contains step by step instructions for integrating Java application servers like Jetty HTTP Server and Orion Application Server Requesting a Module from Quest You receive a module that you add to your Support JProbe installation directory In the Create Edit Settings dialog box your application server and or version is added to the list of available application servers The disadvantage is that this service is not guaranteed Configuring JProbe to Analyze Your Application 65 What Can I Do if My Application Server Is Not Supported Option Advantage Disadvantage Modifying an Application Server You can copy and edit your startup script Startup Script yourself This is probably the easiest way and is most likely to be correct The disadvantage is that you cannot access this configuration from the Create Edit Settings dialog box Creating a Configuration for an Unsupported Application Server You can use the Java Application option to create a configuration for an unsupported application server Before you begin you need to have the following information The path to the main class of your application server The path to the supported JVM you want to use Any arguments you use on the command line The names of packages used in your application The type of analysis you wa
114. the cell and edit the text e To enter the options one at a time in the table double click an empty cell and enter an option 3 Repeat for additional options 38 JProbe User Guide Java Options Enter Arguments As Command Line Text xine Parse Arguments Enter Arguments As One Per Line Xint 4 To rearrange or delete options from the table click a row and use the buttons provided 5 Click OK If there is a mismatch between the contents of the text box and the table JProbe prompts you to select which set of options to use The default is the table The options are listed in the Java Options text box Specifying Your Code You can specify your code during the wizard on the Specify Your Code screen or afterwards in the My Code tab The more JProbe knows about your application the smarter it can be about presenting your data to you JProbe takes the list of packages classes and or methods that you provide that is those set to include and creates a category for your code You can define a Category by specifying its name in the Category Program Name text box and its filters in the My Application Filters table When JProbe connects to an application it looks for a Category with the specified name If a category with the specified name does not exist JProbe creates one automatically with the filters you used to specify the code For more information see Creating and Managing Categories on page 432
115. the method in the Call Graph Call Tree or Method List right click and select Show Source The Source view opens with the source file containing the method code displayed Line data appears in the columns to the left of the method Collecting Data Using Triggers Triggers allow for automated data collection and for precise control over what is recorded They are also useful when you are investigating a program that executes very quickly faster than you could take a snapshot manually You can take a snapshot automatically with a trigger For more information see Setting Triggers on page 425 Running a Performance Analysis 257 Investigating Your Results Changing How You Measure Time You can measure time by either Elapsed Time default or CPU Time Review the descriptions to determine which one suits your investigation You select the time measurement in the Create Edit Settings dialog box on the General tab Note JProbe is a performance analysis tool not a benchmarking tool Times taken while your program is being analyzed cannot be compared to the performance of your program running alone Elapsed Time Use elapsed time to get a sense of how long your program takes to execute in real time Elapsed time includes pauses such as waiting for I O sleeps and any other pauses in the normal execution of your program Pauses may differ from run to run so you should be careful when comparing snapshots using elapsed time
116. the network server you specified To install the License Server run licensing lt version gt platform For example questlicensing4 2 7 l1nx bin To start the License Server follow the instructions in the License Server documentation 4 Added the license file to the License Server via its Web interface The license file was mailed to the email address provided when you downloaded the License Server To open the home page for the License Server see the Readme file To add a license follow the on screen instructions Tip You can also launch the Web interface from the JProbe Concurrent Licenses tab by selecting View License Usage 5 Installed JProbe on the computer that runs your application See JProbe Installation Guide 6 Installed JProbe on each of the developers workstations See JProbe Installation Guide 7 Ensured that all computers have access to the server running the License Server For help with network issues see your System Administrator Setting the License Server During Installation or from Within JProbe This procedure assumes that you are working from the computer running your application and the JProbe Analysis Engine If you do not have a graphical display see Setting the License Server from the Command Line on page 471 JProbe License Options 471 Concurrent Licensing To set the license server for concurrent licensing 1 On the computer running y
117. the parent called the current method CPU Method The CPU resources used by the current method when called by its parent Elap Method Displays the method time for the current method when called by its parent CPU Cml Displays CPU cumulative time when called by parent Elap Cml Displays cumulative time when called by parent Note Inthe Total row Calls Elap Cml and Elap Method data match the information displayed in the Summary pane respectively Call Cumulative Time and Method Time 330 JProbe User Guide Edges To Children The Edges To Children area displays information about the current method s calls to its children Column Data Description Child CPU The contribution by the child method to the child CPU time of the current method Child Elap The contribution by the child method to the child elapsed time of the current method Calls To Child The number of calls from the current method to the child method This column may contain recursion indicators Method The name of the child method called by the current method This column may also contain one of the following recursive indicators h Child is a recursive node Navigating to this child by double clicking takes you to a method that is located up the calling tree from the current method i ee Child is on path to a recursive node for current method Child Time will not include time from the recursive node
118. the server does not start successfully errors are saved to lt WEBSPHERE_HOME gt Appserver profiles lt PROFILE_NAME gt logs lt SERVER_NAME gt native_stderr out Caution Do not end this script using Ctrl C It leaves your application server in an unstable state Run the JProbe start script with clean argument if an abnormal termination occurs WebSphere Profiles If you want to run JProbe with a WebSphere profile you need to set the Server Bin Directory field in the configuration to the profile s bin directory You do not need to set the profilename lt profile gt option IBM WebSphere Portal Server v6 x When you integrate your configuration JProbe creates a startup script resaves your settings file and also creates a jython script called jprobeUpdate py By default the settings file and the jython script are saved to the same directory do not separate these files When you run the JProbe generated startup script you can specify parameters that you usually pass to the WebSphere startup script Do not however specify the server name on the command line because it is set in the configuration If you do specify a server name it is ignored To choose a different server copy the current configuration change the server name and integrate Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the orig
119. to choose between classes with the same percentage you may want to start with the one that contains the highest percentage of missed lines of code The methods in the class are displayed in the Method Detail table bottom pane You can show more columns by right clicking and selecting Show Hide Columns Welcome amp Coverage Session Snapshot Browser A EF v a Name Missed Classes Missed Methods Missed Lines With Line Data Adventure 33 3 M 45 2 55 1 M 100 0 E demos coverage adventure 33 3 M 45 2 55 1 ME 100 0 100 0 100 0 kitchen Bathroom 100 0 M 100 0 100 0 Bi Room 60 0 ME 44 2 M 100 0 B Adventure 29 3 100 0 Logic 28 6 ME 60 3 M 100 0 RoomList 20 0 M 41 4 100 0 Show Only Methods with Missed li gt v Method Method Missed Missed Lines Missed Lines Total Lines demos coverage adventure Kitchen lt init gt Yes 100 0 Catch blocks not filtered 4 Select a method to investigate A good first candidate would be an interesting method with a high number of missed lines Tip Generally speaking you can ignore the results for lt init gt methods If you were to examine the code you would find that initializations such as InputStream inputStream null are reported as missed This result is due to how the compiler identifies the lines in its line data table For more information see Le
120. to continue the performance analysis investigation You can select the method to investigate in the Performance Hotspots tables and click Investigate to open the Methods view with the method in that row as the focus method If you want to continue the analysis with a method that is not suggested in the Performance Hotspots tables you can type a filter into the text box provided in this section and click Investigate In the case that multiple methods match this filter the Matching Methods dialog box opens displaying all the methods that match the filter From here you can select a method and open the Methods view Note The Console remembers the last five filters and displays them in the Investigate list JDBC Tab The JDBC tab is populated with information related to JDBC connections and statements From this tab you can navigate to subsequent views which provide detailed information about methods that offered the connection or statement information Important This tab is displayed only if JDBC data is recorded in the snapshot The Connections section presents the connections created using one of the java sql Driver or java sql DriverManager getConnection 290 JProbe User Guide methods The following table lists the type of information displayed in the Connections table Parameter Information Provided Database URL Represents the URL connection string or if no URL was provided the class of the driv
121. tree For more information see Isolating Subtrees on page 319 Show hide table columns Show Hide Columns Select the columns to show and click OK The selected columns are displayed The rest are hidden Show hide all columns displaying Elapsed Times Show All Elapsed Columns All columns displaying Elapsed Times are displayed hidden Show hide all columns displaying CPU Times Show All CPU Columns All columns displaying CPU Times are displayed hidden if this information is available Hide the column currently selected Hide Column lt column name gt The column currently selected that is the column based on which the Call Tree table is sorted is hidden Exploring Performance Analysis Views 315 Exploring the Methods View Task Resize column widths Action select a node in the Call Tree tab right click and select To set the width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit Response The columns are resized Sort the table by a metric Click the column heading for the metric The table is sorted To reverse the sort order click the heading again Exploring the Method List The following table lists the operations that you
122. true 1 2 Statements DML Only by Elapsed Cumulative Time SQL Method Calls Elapsed Cumulative Time ms SELECT num addr FROM location ORDER BY num java sql Statement executeQuery java lang String 1 127 update location set num addr where num java sql PreparedStatement executeUpdate 2 0 insert into location values java sql PreparedStatement executeUpdate 2 28 Methods Report This report contains the content of the lower table in the Methods view Before generating this report sort the table by the metric in which you are interested Reporting Performance Analysis Results 347 Method Detail Report Methods by Elapsed Method Time Tonija ang Smg dstremerr execute orp apache cerby mpljdbc EmbedConnection presarestatement Garang Sting Method Detail Report This report contains the content of the Edges From Parents and the Edges to Children tables in the Method Detail view Before generating this report sort the table by the metric in which you are interested Parents by Cumulative Time Method CallsTo Cumulative Time ms _ Method Time ms demos performance polynomial Polynomial main java lang String _ J i 3 201 3 143 Children by Child time Method Child time ms Calls To Child demos performance polynomial Polynomial evaluate int int double double 58 751 Performance Difference Report This report contains the content of the Performance Difference view Bef
123. will match the filter and so the filter is not effective The following table contains common syntax and describes how JProbe interprets it If you type this Canonical formis And filters are applied to this ic All methods in all classes in all packages including the unnamed package a All methods in all classes in the unnamed package only C C All methods in classes named C in any package C c All methods in the class named C in the unnamed package only M M Method in all classes in all packages P String P String All methods in any class in the package P whose name starts with String and all methods in any class in a package or subpackage of P String matches both P StringArt draw and P Stringent Check English spelling Reporting Coverage Analysis Results 407 Snapshot Browser Report If you type this Canonical formis And filters are applied to this P C do no change Any method whose name starts with do in any class named C in any subpackage of P except P directly matches P SubPkg C doGet but does not match P C doNotMatch Pre Pre All methods in all classes starting with Pre in any package as well as any class ina subpackage whose name starts with Pre matches both Pkg Prepare m and Pkg Pretend C m s y s y All methods in classes whose name ends with y and either starts with s or belongs to a
124. with JProbe Pre Defined Demo Configurations When you install JProbe the following pre defined configurations are installed and appear in the Configurations Navigator under e Java Applications e Demo Adventure Coverage Analysis allows you to run a Coverage analysis using the Adventure demo e Demo Diner Performance Analysis allows you to run a Performance analysis using the Diner demo e Demo Network Memory Analysis allows you to run a Memory analysis using the Network demo e Apache e Demo Gamepack Coverage Analysis allows you to run a Coverage analysis using the Gamepack demo e Demo Gamepack Memory Analysis allows you to run a Memory analysis using the Gamepack demo e Demo Gamepack Performance Analysis allows you to run a Performance analysis using the Gamepack demo The demo analysis settings file and the JProbe startup script files are by default loaded from the lt JPROBE_HOME gt demos settings directory These pre defined configurations cannot be modified or re integrated if you click Edit Delete or Integrate the Configuration Warning dialog box opens notifying you that the operation is not allowed You can create similar configurations by copying the configuration and modifying it as necessary Copying a Configuration You will likely have more than one configuration It may be easier to start from a copy of a similar configuration than to begin from scratch When you copy a configuration
125. would normally pass on the command line For example some applications can be set up to accept user input data from a file specified on the command line instead of the GUI or the keyboard 34 JProbe User Guide To add an application argument 1 2 Click the browse button beside the Application Arguments text box Do one of the following e In the text box enter the arguments as you would on the command line and click Parse Arguments The table is populated with a list of arguments If an argument was parsed incorrectly double click the cell to edit the text e In the table double click an empty cell and enter an argument Repeat step 2 for additional arguments Application Arguments Enter Arguments As Command Line Text fix Parse Arguments Enter Arguments As One Per Line Fix To rearrange or delete arguments from the table click a row and use the buttons provided Click OK If there is a mismatch between the contents of the text box and the table JProbe prompts you to select which set of options to use The default is the table The arguments are listed in the Application Arguments text box Classpath Add all the paths that are required by your application in the Classpath field as you would if you were using the classpath option on the command line which is essentially what JProbe does behind the scenes It is not a good idea to rely on the CLASSPATH environment variable because it
126. you have a very large program if you are using an application server or if you Know what class of objects you want to track consider using Record Data as Specified in the Table Below because Record Allocations and Stack Traces For All Instances may have a severe performance and memory impact 112 JProbe User Guide Select this option Record Allocations Stack Traces and Garbage Data For All Instances In this situation You want to find what methods are allocating short lived objects in your program or where these objects were allocated that is you want to use the Call Traces or Merged Allocation Points views while investigating by Dead Count or Dead Memory Note If you have a very large program if you are using an application server or if you know what class of objects you want to track consider using Record Data as Specified in the Table Below because Record Allocations Stack Traces and Garbage Data For All Instances may have a severe performance and memory impact Record Data as Specified in the Table Below You know the class or classes of objects that you need to track This option can be particularly useful with large programs where collecting detailed trace or garbage data for all instances is slow and uses a lot of memory Note If you are trying to use one of the other recording options other than Data Recording Off and your program is very slow or you receive errors from JProbe about b
127. 1 0 0 0 33 6 Perm Gen fl 10 1 12 6 25 67 1 Session Connected Finishing current recording and taking snapshot Snapshot C Quest_Software JProbe_8 3 snapshot_raw_1 jpp successfully written Snapshot C Quest_Software JProbe_8 3 snapshot_raw_1 jpp has been added to the snapshot navigator as snapshot When you run a session the Memory Pools chart displays the available size of the Java heap and the memory in use by your program over time You can set the time span for the graph For more information about this view see Exploring the Runtime Summary View on page 274 222 JProbe User Guide Understanding Performance Data This chapter helps you to gain a better understanding of how data is captured by the Performance analysis tool and displayed in the Methods view It also introduces the features that you can use to fine tune your analysis and your results This section covers the following topics e Collecting Data on Methods e Building a Calling Context Tree e How Data is Affected by Using Performance Features e Recursive Methods Collecting Data on Methods The Performance tool captures information on how long methods and lines of data in your application take to process In order to perform this function JProbe needs to know when methods are called and when they are returned JProbe relies on its own instrumentation to generate these events which it does by interfacing with the JVM and if permiss
128. 207 4 968 My Code o a Method Count 0 4 968 0 Method Memory 0 4 968 0 208 JProbe User Guide Merged Allocation Points Report This report lists the locations of all the instances in the Merged Allocation Points view in order of cumulative count If you are investigating by Dead Count Memory these are the locations of garbage collected instances otherwise they are the locations of live instances Before generating this report sort the table by the metric that you are interested in Locations by Cumulative Count Tome Cumulative Cumulative Memory Count Bytes demos memory sim Sim run 5 000 80 000 java lang Thread run 5 000 80 000 Pe a la java lang String java lang String 5 000 80 000 demos memory sim Sim Client start 5 000 80 000 Heap Graph Report This report contains the contents of the table in the Heap Graph view Before generating this report sort the table by the metric that you are interested in Heap Graph by Referrers Name Referrers References In Use Case javax swing plaf ColorUIResource OX8caf 67 O false sun swing PrintColorUIResource 0X972b 41 1 false java util Locale 0X7c26 27 3 false javax swing plaf InsetsUIResource 0Xa432 25 0 fase sun awt AppContext OX8b80 23 5 fase javax swing plaf FontUIResource 0X8b71 21 2 false java awt ComponentOrientation 0X8cc7 20 0 fase javax swing plaf BorderUIResource CompoundBorderUIResource 0Xa35f 14 2 fase javax swing plaf m
129. 248 your results 243 isolating Call Trees 228 JDBC component instrumentation 267 measuring improvements 258 Memory Pools chart 221 275 Method Detail view 325 Methods view 297 narrowing your investigation 249 no encapsulation 228 pruning Call Trees 228 public methods 270 recursive methods 232 reports 343 running a basic analysis 235 a session 237 Runtime Summary view 274 saving baseline snapshots 239 session runtime tips 351 Set Recording Level 236 setting up JProbe for an analysis session 236 snapshot saving a baseline 239 Source view 335 strategies for program efficiencies 218 Summary view 283 Threads 276 482 JProbe User Guide timing settings 266 triggers 271 troubleshooting tips 350 understanding data collection 240 undo and redo buttons 323 views setting preferences for 342 Performance Difference view interacting with 340 opening the 338 understanding the 339 Performance Methods view interacting with 305 opening the 297 understanding the 297 Performance Runtime Summary view interacting with 279 understanding the 274 Performance Source view interacting with 336 opening the 335 Performance Summary view interacting with 292 opening the 283 understanding the 285 performing a basic Memory analysis 112 Pinned Instances report 205 Pinned Instances view interacting with 157 opening the 155 understanding the 155 port number changing 48 portable heap dump importing 127 preferences setting 453 printing to PDF 441
130. 4 H3 H4 H5 v5 Unfiltered Call Tree During filtering the call tree is mostly preserved but there may be some methods that are not exposed For example V3 may have called a method that called another method that eventually called V5 Full Option Tip The Full option requires a lot of resources during the collection of data impacting the amount of time it takes to run a session Choosing the Full encapsulated option provides the most comprehensive way of viewing method information The structure of this call tree is most like the call tree representation of your application data The data shows only the first encapsulated method all other information about encapsulated methods is removed from the call tree 252 JProbe User Guide Root Main v1 _ v2 5 amc vs Encapsulation using Full option Notice in the Unfiltered Call Tree on page 251 that the nodes H4 and H5 are displayed in the call tree but when using the Full encapsulated option the nodes H4 and H5 are encapsulated and not shown The time spent in these nodes is directly attributed to their parent methods Grouped Option Tip In comparison to the Full encapsulation option the Grouped option does not substantially impact on performance The Grouped encapsulated option aggregates all of the encapsulated methods called by a particular method into a single method or node This type
131. 60 JProbe User Guide all the details are preserved You can then make edits and save the configuration under a new name To copy a configuration 1 Select the configuration in the Configurations Navigator 2 Right click the configuration name and select one of the following options Copy settings for Memory Analysis Copy settings for Performance Analysis Copy settings for Coverage Analysis A copy of the original configuration appears in the Configurations Navigator and the configuration settings are displayed in the Configuration Details tab 3 To edit the new configuration click Edit The Configuration Details tab is placed in editing mode a b c Type in a name for this new configuration Edit the settings on any of the tabs as necessary Click Save 4 To generate a startup script click Integrate a b c d Type in a location and file name for the settings file and click Save Click Continue Type in a location and file name for the startup script and click Save Click Close Wizard Editing a Configuration When you edit a configuration the settings are saved to the existing settings file In most cases you do not need to reintegrate your configuration To edit a configuration 1 Select the configuration in the Configurations Navigator Configuring JProbe to Analyze Your Application 61 Managing Configurations 2 Inthe lt Settings_File_Name gt tab click Edit or Right click the configura
132. AG ili A AA A EAE AT 400 lasg Summa N a taste east aise aia a ecu aaa a oe 400 Class and Method Summmaries ssessessesssessesssessecsessecssssnesatsentsasessseceesaecssesseestesseesseestecseesaeseestecneesseentesaeens 400 Snapshot Browser Reporte tiarren i eiii A di NEE N E N i 401 Application SUMMAN senesi eso eee NEENA TEKER AN 402 Package SUMMALY ieie anal Goin RE ER E NEAS 403 ESAE A E EA A E 403 BEOU o A D Trs PAPE caster ag EA RT 404 Report Filters iisiscctis ccsie dived alee O a E Galas 405 Troubleshooting Coverage Analysis ReSUItS ssssssssssssessssssssesssesessseeeseeseesesseeeesceeeeneeneenesneesseneeteaeeees 409 Coverage Troubleshooting Tips Learning How to Use JProbe Effectively Hints for Success with JProbe sissies es TE R E Ei Understanding the Architecture for Your Program sss ssssssssssresssrissssressssrrnssnreesnrnnsnnrnnnnrnennnrnnnnrtennrennnnrnennree 414 Selecting USE Oa a araea ap A E aE EEEa TA AEE AAT EAA A AA Ea Compiling Your Program Leaming How JProbe WorKSiiissrarinona nran ROEN adenine ae JProbe Analysis ENJINS iussi n aaa a ele nine ie JProbe Connection Managers srtcsawideniinenie aealoiaaa die N E anh aaae 417 JProbe Console iseisicssisecde EA AEST AT Probe Snapshots a a Ea cat aaa Aiea lta ae Demos and Tutorials arnan in aaa AR aaae ia saaa buries hhecebasinnasdnr anes A Quick Tour of the Interface Configurations Navigator Snapshots Navigatora e
133. All methods in any class in the package P whose name starts with String and all methods in any class in a package or subpackage of P String matches both P StringArt draw and P Stringent Check English spelling Any method whose name starts with do in any class named C in any subpackage of P except P directly matches P SubPkg C doGet but does not match P c doNotMatch All methods in all classes starting with Pre in any package as well as any class ina subpackage whose name starts with Pre matches both Pkg Prepare m and Pkg Pretend C m 144 JProbe User Guide If you type this Canonical formis And filters are applied to this s y s y All methods in classes whose name ends with y and either starts with s or belongs to a subpackage that starts with s matches both java rmi server RMISocketFactory com quest say and java security cert X509CRLEntry foo bar foo bar foo bar foo bar String t e foo bar no change All methods in the class bar in package foo This is likely not what you wanted See the next example All methods in all classes in foo bar or any of its subpackages Methods whose names start with t and end in ein foo bar String matches both toUpperCase and toLowerCase Interacting with the Instances View The following table summarizes the tasks that you can perform Task See the instances that dominate the selection OR the li
134. AnalySis cscsssscsssssecsesseesesesseeeesesseeceecseeseecsneeseaceeeseeseeseaeeneeseateeeseneeeeseaneaeeseaes 235 Setting Up JProbe for a Performance Analysis Session s sssssssssssessessesseseessessessessseussseesessessteaeeseeansateateaneaneetsaeess 236 Setting the Recording Level ciii c ccciscsccsiis nc sotevccscleveci sac vcssseveassliveesciudevedsetveseeesdis E EEEE E EAE E E EA 236 Running an Analysis Sessions iscsi netiientiwsnths e a a a 237 Exercising Your Program sirsiran ritan anenai aaa Ea aupian ciate EEAS EATER NEENA EAS aia 237 Saving Baseline Snapshot n eee tte teeter d 239 Understanding Data Collection n sas iueve cde nriinaat a a a aan aa nE aAa 240 Investigating Your Res lSi diii iie E aden E A T alan ea 243 Identifying Performance Bottlenecks ciccec 2cacecsa sens arena daer nari eericnenannarewann 243 Investigating Performance Bottlenecks ssesesessseseesesstesseeseeseeseeatestesteasestesnsenssssessesseensenteateneeateateaneateaneess 248 Narrowing Y our Investigation siti a 2icuaaieada eva da A andra 249 Measuring Performance Improvement ssscesssscsesssssessessessesseesesaseseeseseseseesseseeaeeneseessteneesessteasesteaeeateaensees 258 Identifying Deadlocksi a0 tose eben aie ce ie rg ee ea ee E 260 Investigating Performance DeadlockS cssesssesesesesssssesseesseseesteatesteseeesessenseessassseassateneeateateneenteateaneateaeens 264 Learning More About Performance Options s
135. CSV File on page 455 Create a report for this view Click Export Data to HTML a The Create Call Traces Report dialog box appears For details see Call Traces Report on page 207 Exploring Memory Analysis Views 181 Exploring the Merged Allocation Points View Exploring the Merged Allocation Points View The Merged Allocation Points view displays a merged stack trace for a set of instances The merged stack traces lead you to the instance that allocated many objects over the runtime of the program By default the stack traces are collapsed and only the allocating methods appear In order to view merged stack traces you need to run the session with the Collect Stack Traces option chosen Note This view is not available if you have set the Investigate By option to Heap Count or Heap Memory in the Instances view For more information see Understanding the Instances View on page 139 This section contains the following information e Opening the Merged Allocation Points View on page 181 e Understanding the Merged Allocation Points View on page 181 e Interacting with the Merged Allocation Points View on page 182 Opening the Merged Allocation Points View The Merged Allocation Points view is one of the views where you can focus on a class selected in the Instances view To open the Merged Allocation Points view 1 Inthe Instances view select Recorded Count or Recorded Memory
136. Click the up arrow on the Coverage Bar or right click and choose Previous Stripe The previous block of code missed lines or missed conditions is displayed Navigate to a method within the code Choose a method from the Method drop down list at the top of the Source view The method is displayed Examine a missed condition in more detail requires conditions data Choose a missed condition in the source code Information about the chosen condition is displayed in the Condition Detail table Color code source by missed lines and or missed conditions requires conditions data Beside Show Data For choose Lines Conditions or both checkboxes The source code colors are adjusted accordingly Show hide columns Right click and choose Show Hide Columns Choose the columns to show or hide and click OK The chosen columns are displayed The rest are hidden Rearrange columns Drag a column by its column heading and drop it in a new location The column appears in its new location 390 JProbe User Guide Task Resize column widths Action e To set the width drag the edge of the column heading left or right e To size a column to fit its data right click the column and choose Size lt column gt to Fit e To size all columns right click and choose Size All to Fit Response The affected columns are resized When sizing all columns to fi
137. Consider the following example A doGet B appServerImpl X1 append In the Call Graph X1 append has 5 nanoseconds of method time but we cannot tell how much of this is due to the code of the application By isolating A doGet a copy of the CCT is created whose top node is A The data for main B and all of B s children including the calls to x1 down that part of the tree is discarded The data appears the same in the CCT or the Call Graph The isolated tree contains only the times and counts associated with the doGet method This is just the branch of the CCT rooted at A CCT rooted at A The node x1 has time of 3 nanoseconds and 2 calls as a result of the application code For more information on isolate see Editing the Data Set on page 318 Strategies for Performance Analysis 231 Understanding Performance Data Pruning Subtrees Pruning removes part of the CCT and therefore changes the shape and alters the values in the graph This command actually removes data from the Call Graph Note This feature differs from the Hide Node option on the graph which only affects visibility and not the values in the Call Graph For example in your profiling results you see class loading activity There is little value in viewing the data for the class loader because you cannot modify its code to improve its performance By selecting the loadClass method and pruning it the time spent loading it woul
138. Creating and Managing Categories Manage Categories To create a category or folder select the folder that will become its parent and press New Folder or New Category Use drag and drop to move or reorder categories Sale me Category Program Name Java SE S L Root Categories S F System Code Parent System Code ava gi Filter List EE Java Tools EE Srd Party Frameworks Category or Pattern Action ce fil My Code Java VM Threads include Other Java SE include Java Base Libraries include Java Integration Libraries include Java UI Toolkits include Primitive Arrays include New Folder New Category Restore Defaults ok Cancel Help Select the folder in which the category logically belongs and click New Category Enter a name in the Name field Tip Itis recommended to choose a Category Name that is descriptive for the code you will be looking at It is also recommended to specify unique Category Names since overlapping patterns in Categories can be confusing and you can use the same Category Name and filters in multiple configurations To add packages to this category e Click the browse button in the cell navigate to a custom package and click Open e Double click the cell and enter the package name The package is displayed in the cell and the Action is set to Include To exclude some code from this category click in an empty cell
139. E QUEST SOFTWARE JProbe 8 3 User Guide 2009 Quest Software Inc ALL RIGHTS RESERVED This guide contains proprietary information protected by copyright The software described in this guide is furnished under a software license or nondisclosure agreement This software may be used or copied only in accordance with the terms of the applicable agreement No part of this guide may be reproduced or transmitted in any form or by any means electronic or mechanical including photocopying and recording for any purpose other than the purchaser s personal use without the written permission of Quest Software Inc If you have any questions regarding your potential use of this material contact Quest Software World Headquarters LEGAL Dept 5 Polaris Way Aliso Viejo CA 92656 www quest com email legal quest com Refer to our Web site for regional and international office information JProbe Patents Patent pending Trademarks Quest Quest Software the Quest Software logo AccessManager ActiveRoles Aelita Akonix AppAssure Benchmark Factory Big Brother BusinessInsight ChangeAuditor ChangeManager DeployDirector DirectoryAnalyzer DirectoryTroubleshooter DS Analyzer DS Expert ERDisk Foglight GPOADmin Imceda IntelliProfile InTrust Invirtus Token I Watch JClass Jint JProbe LeccoTech LiteSpeed LiveReorg LogADmin MessageStats Monosphere NBSpool NetBase NetControl Npulse NetPro PassGo PerformaSure
140. EF to open the Create Summary Report dialog box Select a location for saving the report and click Generate File Response The report is generated in the specified location Exploring the Methods Tab The following table lists the operations that you can perform in the Methods tab Task Move to the Methods view with a focus method Action Double click a row in the Performance Hotspots tables OR Select a row in the Performance Hotspots table right click and select Investigate OR In the What Next section type in a method name in the field or select one from the list and click Investigate Response The Methods view opens with the selected method as the focus method Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized 294 JProbe User Guide Task Action Response Export the table to Right click in the table and The file contains the table CSV select Export to CSV Type a data separated by a file name and click Save comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Exploring the JDBC Tab Important This tab is displayed
141. HTTPS JMX Remoting Connector 61613 a 0 0 0 ActiveMQ Transport Connector 61616 0 0 0 9 ActiveMQ Transport Connector Started Application Modules EAR gamepackear gamepack ear 1 car RAR org apache geronimo conf igs system database 2 1 4 car Web Applications gamepack Geronimo Application Server started 4 Wait for the startup activities to finish 5 To monitor and or interact with the analysis session a Launch the JProbe Console on your local computer b Click Attach to Session f on the toolbar JProbe searches for currently running engines and JVMs and displays this list in the Attach to Running Session dialog box i Attach to Running Session Probe Engines JMX Select the engine or enter the Host Name and Port Number below the table Name Product User com sun enterprise glassfish bootstrap ASMain Memory azafar Memory Memory lupas Host Name IP Address localhost Port 52991 x OK l Cancel Refresh Starting Your Application with JProbe Attached 75 Runtime Notes for Application Servers c Inthe JProbe Engines tab select an engine from the list of currently running engines or type in the Host Name IP Address and Port fields the host name and port number of the computer running your application under JProbe Tip The computer may be localhost or a remote computer The port number must be the one specified in the JProbe Execution Console w
142. Panis Rapant assasisosenaianaaaa 208 US SRA TAIT PRECIO RA E AE eensaenasisntvicasesesetlareiiaidnanareergratiaatens 208 Memory Snapshot Difference Report csccsssesescseseesesseetesseeeeseeseeseeseesseetesnsetsneeateneeateneeass 208 204 JProbe User Guide Setting Report Options When you generate a Memory report the number of instances in the report is controlled by the Number of Rows to Report option in the Create Memory Report dialog box The default is the top 10 Create Memory Report Output File Reports LeakExample_Runtime_by_Heap_Count Number of Rows to Report After sorting report top 10 v Generate File Cancel Preview _ Help Runtime Summary Report This report contains the contents of the Data tab in the Runtime Summary view Before generating this report ensure that the data is current by clicking ee Refresh Runtime Data and sort the table by the metric in which you are interested Runtime Instances by Heap Count Nome Use Case Use Case Memory Heap Heap Memory Dead Dead Memory Count Bytes Count Bytes Count Bytes char 462 23 112 5 280 305 728 0 0 java lang String 223 5 352 5 038 120 912 0 0 short 2 32 1 746 228 520 0 o java awt Rectangle 1 431 34 344 1 533 36 792 0 o byte 95 11 664 1 485 305 712 o o int 2 544 1 408 144 200 0 0 java lang Class 0 oO 1 288 548 784 o o java util HashMap Entry 49 1 176 968 23 232 0 0 java lang Object 442 11 088 867 33
143. Probe the wrapper startup batch file and wrapper conf configuration file The batch file does not require any modification It is used only to start the application with JProbe The integration process creates a new wrapper conf file with modifications to integrate the Java Application with JProbe Later this file is passed as a command line argument to the startup batch file for starting the server with JProbe If you want to start the application server on a command line window with JProbe you must provide the path to the JProbe configuration file as command line argument to the script file Example e for MS Windows App bat lt complete path to JProbe s wrapper conf file gt e for Unix App sh script lt complete path to JProbe s wrapper conf file gt Oracle AS 10g v10 1 3 0 or below When you integrate your configuration JProbe creates a startup script saves your settings file and also makes a copy of the Oracle Application Server s opmn xml file called lt startup_script_name gt _opmn xml In lt startup_script_name gt _opmn xmil java bin data value is replaced with the path to the jplauncher executable Caution You should make a backup copy of the original opmn xm file If you encounter a problem you can manually restore and load your copy of the opmn xm file Usage lt configuration_startup_script gt Starting Your Application with JProbe Attached 91 Runtime Notes for Application Servers The startup
144. Run dialog By default the startup script has the same name as the configuration The JProbe Execution Console window is displayed It contains the port number that JProbe is using for this session JProbe Execution Console demos memory leakexample LeakExample x Probe version 8 3 6 copyright 200 Quest Software Inc ALL RIGHTS RESERVED Probe is a registered trademark of Quest Software Inc lemory analysis started Lb Probe Connection Manager is available for Console connections on port 52991 2 To monitor and or interact with the analysis session a b Launch the JProbe Console on your local computer Click Attach to Session f on the toolbar JProbe searches for currently running engines and JVMs and displays this list in the Attach to Running Session dialog box i Attach to Running Session Select the engine or enter the Host Name and Port Number below the table Name Product User com sun enterprise glassfish bootstrap ASMain Memory azafar Memory Memory lupas Host Name IP Address localhost Port 52991 v OK l Cancel Refresh Starting Your Application with JProbe Attached 73 Running a Java EE Application with JProbe c Inthe JProbe Engines tab select an engine from the list of currently running engines or type in the Host Name IP Address and Port fields the host name and port number of the computer running your application under JProbe
145. SHALL QUEST BE LIABLE FOR ANY DIRECT INDIRECT CONSEQUENTIAL PUNITIVE SPECIAL OR INCIDENTAL DAMAGES INCLUDING WITHOUT LIMITATION DAMAGES FOR LOSS OF PROFITS BUSINESS INTERRUPTION OR LOSS OF INFORMATION ARISING OUT OF THE USE OR INABILITY TO USE THIS DOCUMENT EVEN IF QUEST HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES Quest makes no representations or warranties with respect to the accuracy or completeness of the contents of this document and reserves the right to make changes to specifications and product descriptions at any time without notice Quest does not make any commitment to update the information contained in this document License Credits and Third Party Information See Third_Party_Contributions htm in your JProbe doc installation directory User Guide November 2009 Table of Contents Introduction to This Guide arar a err ra rr araea ar oet aaao a aeaeo K arora aaraa ra AEE eiaa 15 About JPrODE zinin opaa AA AE E AE a a E A a R E 16 About This GQuid osni aa a A A EN N EELA A 16 JProbe D c mentation S ite masaressennrenririinri ieni AnA en aii EAR 18 Core Documentation Setinin ise iiie EERE NAE RENEE 19 Feedback on the Doclmentation s s 2a8i tat shai hieh nai adel a 19 TXT CONVENT ONS e E a a aa E a r Aa a EE RARE a EAA AA ERT tty 20 About Quest Software INC a A AN 20 Contacting Quest Somwaresesii s cacccess haces beacestsesases hiaccidstassateraassisetrssatansansaneamsega cia cot aR EEEa p caine Ta Ge
146. Subtrees Exploring Performance Analysis Views 319 Exploring the Methods View e Pruning Subtrees Isolating Subtrees You can manually edit the data set by isolating a call tree For example if you want to focus on a single thread or the subtree of a method you isolate the tree in the Call Graph The method is anchored to the root node and the data set is recalculated to represent only that tree For examples of when you might use isolate see Focusing on Threads and Methods in the Call Graph and Call Tree on page 249 The following diagrams show a Call Graph before and after a subtree is isolated The dark nodes represent the target subtree Original Call Graph After Isolating Subtree Note Line level data is not recalculated While the Call Graph displays the newly recalculated method level data the Source view continues to show the original line data Pruning Subtrees You can selectively edit the data set by removing call trees from the Call Graph Essentially the graph is a set of superimposed subtrees each representing a series of calls You can remove layers of subtrees by pruning them from the graph The data set is recalculated to eliminate the contribution of the removed methods For examples of when you might use prune see Removing Threads and Method Subtrees on page 250 In the following example the dark nodes represent the subtree to be pruned away Notice that methods that were also called by
147. Text Select a new font from the Name list Select Regular Bold or Italic from the Style list Select a font size from the Size list Your choices are displayed in the sample text below the list boxes Click OK Interacting with the Call Traces View The following table summarizes the tasks that you can perform Task Action Response Remove a selected Select a method and click The view is updated to node and its Ys reflect the removal of the children from the Prune selected node and its dataset children from the dataset Exploring Memory Analysis Views 179 Exploring the Call Traces View Task Limit the dataset to a selected node and its children Action Select a method and click Isolate E Response The view is updated to display only the selected node and its children Show the stack frames contained in the selected node Select a method and click Replace Category with More n ra Detail The view is updated to expand the data in this view Investigate instances allocated by a method Select a method and click Instance Detail IE The Instance Detail view opens to display the instances allocated by the method Expand the tree to show multiple paths Select a node right click and select Expand To Next Branch Point The node expands to show more than one branch View the source code for an allocating method Select a method right c
148. The selected node and its subtree are removed from the graph All metrics are recalculated without the contribution of the pruned call tree For more information see Pruning Subtrees on page 319 Isolate a subtree Isolate Selected Node The selected node is anchored to the root node All nodes that do not belong to the method s subtree are removed All metrics are recalculated to include only the data collected on the methods in the isolated tree For more information see Isolating Subtrees on page 319 Re draw the graph Redraw Graph Based On Resets the graph clears using the select node Selected Node the history and re draws as the focus method it using this node as the focus method Move to the Method Open Method Detail View The Method Detail view Detail view with a opens with the selected focus method method as the focus method Exploring Performance Analysis Views 311 Exploring the Methods View Task View the list of methods represented by this node Action select a node in the Call Graph tab right click and select View Method Names Response Opens the Methods dialog box which contains a list of method names represented by this grouped node Drill down to source code Show Source If you have the source code the Source view opens with the source code displayed Drill down to line data Show Line Data The Source view associ
149. Time and the Color to Show metric is set to Hot Paths Elapsed Time 5 If you are analyzing a Java EE application you can focus your investigation by isolating the tree that represents your application Find the entry point to your application for example a doGet or doPut method select a method right click and select Isolate Selected Node 246 JProbe User Guide 8 9 The data set is recalculated to include only the methods called by your application To zoom out so that you can see the critical path more clearly move the Scale slider to the left Cumulative time includes the time spent in child methods Therefore your Call Graph is brighter on the left side where your root nodes appear You should also see a bright path of nodes through the Call Graph from left to right This is your critical path welcome AE Summary Methods 7 Color to Show Hot Paths Elapsed Time v LN a ish a fal 5 eae J Call Tree Call Graph Follow the critical path from left to right expanding the call tree as necessary and select the last bright node that represents one of your methods For details see Finding Hotspots by Navigating the Call Graph on page 247 Zoom back in on the selected node by sliding the Scale button to the right Position your mouse cursor over the selected node in the graph A tooltip shows the method name and signature the cumulative time and the number of calls to the
150. Understanding Performance Dele i c cidicciccnidcdicnnianncannincninanananan 218 JProbe User Guide Strategies for Achieving Program Efficiency There are two common ways to use JProbe Performance e Analyzing an Existing Program e Focusing on a Known Inefficient Algorithm Analyzing an Existing Program Senior developers often use JProbe Performance at the end of a development cycle to improve program efficiency JProbe Performance highlights hotspots in your program which helps you to identify performance bottlenecks algorithms that are too slow or that block other processes for too long To analyze the entire program you should be an experienced software developer with good overall knowledge of the architecture and expected behavior of the program Strategies for Performance Analysis 219 Strategies for Achieving Program Efficiency Program Use Cases Architecture Hypothesis Set Method Level Detail Filters on Your Packages 1 Add a Line Level Detail For analyzing a known I Filter on the Method inefficient algorithm only Containing the Algorithm You may want to run your Set a Baseline for the session a few times to allow Use Cases the JVM to perform optimizations Locate and Fix Performance Bottlenecks Do the results match your hypothesis Measure Improvements Against the Baseline While not required you should take some initial performance snapshots of your progra
151. Y Leak Doctor Instance java lang Class Ox2df Name comment type Fede S Keep Alive Size CONSTANT_POOL ARRAY FIELD FIELD lt class sun reflect GeneratedMethodAccessor2 java lang Object 0x2636 java lang reflect Method 0x261e java lang reflect Method 0x2622 E java lang reflect Method 0x261e OB8AaBAa lt class javax swing UIDefaults 5 FIELD CONSTANT_POOL 7 SS Sb lt constant pool 6 gt 912 java lang Object 0 56 clazz 2 616 clazz 80 constant pool 188 gt 824 sclass gt sun reflect misc MethodUtil CONSTANT_POOL lt constant pool 167 gt 3 352 lt class sun reflect misc MethodUtil 1 CONSTANT_POOL lt constant pool 34 gt 344 sun reflect misc MethodUtil 0x3 CLASS lt class gt sun reflect misc 880 a Sticky Class Root 0x43f8 From Root ROOT This JavaNameFormatter 0 Name _ Comment Type _ Field Name Keep Alive Size F Root Ef Static Root y Recording Boundary BB cycle The upper pane of the Leak Doctor lists references to the loitering object by name optional comment type of reference and the field of the referring instance that refers to the object The toolbar has buttons for removing and restoring edges The lower pane displays the removed edges 192 JProbe User Guide The name of each field Leak Doctor tree is prefixed by an icon which provides information about that field as described in the following table
152. _Memory_WebLogic10 Integration status Checking JPL file existence done Saving current settings in JPL file done Validating WebLogic server script file done done The selected startup script file contains no call to java executable Creating JProbe script file in directory C Program Files ApplicationServers BEA_Systemsiuser_projectsidomainsweblogic Oigamepackibin Reading startup script startWVebLogic cmd to replace SJAVA_HOME bin java with jplauncher exe done Please save JProbe startup script file to continue Note We recommend that you use the default script directory or original startup script directory to save JProbe startup script file Save JProbe startup script file The startup script is required to execute an application with JProbe Where would you like to save the start with JProbe script file C Program Files ApplicationServers BEA_Systems user_projects domains weblogic10 gamepack startWebLogicWithJProbe cmd Save 52 JProbe User Guide Review any messages For example a configuration for an application server usually includes notes about how to run the JProbe generated startup script Tip Runtime notes are also available from the Help button on the application server s tab available after the integration is completed or under Runtime Notes for Application Servers on page 75 If warnings are displayed you may need to fix some configuration settings Click Cancel and edit t
153. a 21 Contacting Quest SUpport E alii E E EEE 21 Quest COMMUNES iaei ariii ARa ERE N AENEA 21 Configuring JProbe to Analyze Your Application sessssusessusuuurusnunuuuunennunsnuununnunsnnunennununnnnennunununnrnnunnnnnnnns 23 Configuration Qverviw tise seccsesse sees gece acessenaszssazsacaasseuapesstsansachonedatee AA EEE EESAN aE OE ana COASA ia RAA ESEA ARAE AEE i 24 Contiguring awava EF Application n osaa n a ae S E AA A Ae 24 onfiguring a Java SE ApplicatiO M sarees sestanke aaeeea ane a aa aaa EES O REEE a aa AE EA AAAA E ENARE iia 27 Defining a JProbe Configuration issnin EEE EEE EEE E 29 Specifying Your Java EE Application Server cccccscsscsessessessessnssessesssessessssssssessessessnsanssnsansanssnsanseessnsansesenseeseseeseneeeses 31 Specifying Your Java Application ss cecsiesccsscsiccccsas acsetsaaccscideaceetleaesssidecesesseteeesiv ded st enveai teaveesl asveveerdessbbectaateereccubevesiheets Main Class and Working Directory Application Arguments i2ss sizrnsiana a a EA E a AEEA EAE lE ETI oK V EE ETT A A E A AE Specifying a Java Virtual MaChine 2 vices scan anand nan AEE ARO N a 36 Java Executable i cit diareeha ie ieee E EAO 36 Java ODIOM Segnen teed arvieted A N ah dedi ARN SPecitying LOTA Oeo leE E E EE EA ATT A RAE My Application Filters Table 4 JProbe User Guide Elements Table eminen daw wei diene caine a A Specifying a JProbe Analysis Type c sessssssesssecsessessssecsses
154. a 292 Exploring the Methods VIEW nsciis nen n TERA R E N 297 Opening the Methods Viewin E E AA 297 Understanding the Methods View ssssssssesesssessessssessseseeseeseestesessssssesesstssenteaseanenteaneanennsaseeenseateanennenes 297 Interacting with the Methods View secsesssssssessessessseseeseseteseecseetesseesesseesseneeseensesseseeaeeseeseeateseeaseateateseeeesnenees 305 Customizing the Call Graph ees cozcsseisecser veces dines a vanacavea nivenacanertnactesiee AE 323 Exploring the Method Detail VieW c ccecsessssessessesesseseesesseeeeseeseeesneeueseeaseesseeassneaeeesseeassessteasaeeaseneaseesaeeaseneateaseeeaes 325 Opening the Method Detail View 0 seeesestessesssesecssessessecseesneeaecssecneeaesseesseestesseesecseesueeneeseeseesneeseeneeseeens 326 Understanding the Method Detail VieW c esccsccessesssesseseeseeseeeeseesseeeseeesseeaeseeseeesseeasesseeassesseeasseeaseneateasaeeaees 326 Navigating the Method Detail View Customizing the Method Detail View Exploring the Performance Source ViGW eessessessesseesessnesstecsessuesstesneestecseesseeseesseesteesesueesueeateeneeateenteaieeneeaneestennees Opening the Source VEW 22 sicssancesiesnancaran cansaienscnndaetnaniam ature avin lat AE EN Interacting with the Source ViCW ceseessesesstecseesseesecstecsessseestecseestecseesueeseesseestecseesueesueeateenssaeeneeaeenteaneententees Exploring the Performance Difference View sc sccsssecssse
155. a UI Toolkits Primitive Arrays rm Bother Pinned Instances by Type Fiter Classes v t 286 286 Name Recorded Count Heap Count Recorded Memory Heap Memory Keep Alive Size 66 576 202 368 440 61 080 1 440 87 608 3 272 35 480 2 688 27 264 32 25 096 g l java lang String 2 774 G java util concurrent ConcurrentHas 1 G java util HashMap 36 sun java2d loops GraphicsPrimitive 2 sun reflect DelegatingClassLoader 48 l java io PrintStream 1 Investigating Heap Instances G Immediately Pinned The upper pane Dominating Instances by Instance Type contains a table which shows details about the instance set selected in the Instances view Each row represents the set of dominating instances that you selected as input to the view when selecting What Do These Pin option from the list and they are responsible for keeping alive the instances shown in the middle and lower panes The upper pane shows the starting set of instances from which the data in the rest of the view is derived These instances are grouped by class so instead of seeing a row for each instance you can see only one row summarizing the count of instances similar to the Instances view table The middle pane Pinned Memory by Pinning Type shows the categories of instances pinned by the starting set The pane contains a chart which illustrates the percentage of pinned instances by Keep Alive Size in each of th
156. able lists the navigation tasks Task Focus on call path to current method from a parent Action Select parent in Edges From Parent Response The CCT nodes for the current method which have the selected parent are displayed in the CCS table Focus on a call path from current method to a child Select child in Edges To Children The CCT nodes for the current method which call selected child nodes are displayed in the CCS table Exploring Performance Analysis Views 333 Exploring the Method Detail View Task Change the current method Action Double click the data in Edges From Parent or Edges To Children Response Changes the current method back to the selected method Move back in view history Use Back Changes the current method back to the previously viewed method Move forward in view history Use Forward Change the current method to the next method View source code for the current method Click Source 3 to navigate to the source code when available of the selected method Opens the Source view for current method or to the Please Find Source dialog Customizing the Method Detail View The data displayed in the Edges From Parents Edges To Children or Calling Context Sites tables provides you with a better understanding about methods and those methods that are deemed expensive You can customize each tab to display data that is
157. ade some optimizations or rearrangements when creating bytecode For more information see Nonintuitive Results for Conditions on page 396 Missed Column This column simply states if the line of code was missed A Closer Look at the Data in the Condition Detail Table When you choose a condition in the source code the Condition Detail table is populated with details about the condition coverage Switch data If the condition is a switch case the table includes the possible case values including the implicit or explicit default and whether or not each was executed Exploring Coverage Analysis Views 387 Exploring the Coverage Source View Switch data case Ht o Mo 1 Yes Default No Conditional data For non switch conditions the table displays the conditional opcode mnemonics from the line and for each indicates whether execution of the condition resulted in a true or false outcome Conditional data Opcode True False if_icmple Yes No If you are unfamiliar with opcodes you may want to read the following background material Your java compiler for example javac compiles Java source code into instruction opcodes known as bytecode for the Java virtual machine Conditions in the source code are translated into condition opcodes in the JVM instruction set The condition opcodes are like a simple if statement that only ever checks a single true false condition A compound conditio
158. al number of possible condition outcomes in the class package application taking filters into account With Condition Data Percentage of lines of code with condition data Exploring Coverage Analysis Views 381 Exploring the Snapshot Browser For the Missed and Total columns data is rolled up the tree Values for the package represent the sum of all classes in the package while the values attributed to the application represent the sum of all its packages The name of each field in the Application Tree is prefixed by an icon which provides information about that field as described in the following table Icon Type k Application node 3 Package node Class node A Closer Look at Data in the Method Detail Table By default each method states whether or not it was missed the total lines of code in the methods the total number of conditions if available and the name of the class You can choose to show additional columns The following table describes the columns in the method detail table The columns marked with an asterisk are shown by default The other columns can be added to the view by right clicking in the table to open the Show Hide Columns dialog box where you can choose other columns to be displayed Column Name Description Method Name of the method Method Missed States whether or not the method was missed Missed Lines Missed Lines Total Lines Missed Line
159. alize how memory is being used in the heap As you run your sessions you will observe some common behaviors in the chart including a steady graph increasing growth growth and a drop and spikes Note You may notice a growth in the heap when you start recording data with some JVMs This is a result of engaging the JVM s profiling interface You may wish to adjust your heap size to accommodate this growth Steady graph A steady graph indicates that some core set of objects are remaining in memory You need to assess whether this is expected behavior or symptomatic of a problem Refer to your program s architecture for a list of persistent objects A good understanding of 104 JProbe User Guide your long lived objects helps you to determine if they are holding short lived objects longer than necessary Increasing growth Increasing growth in the Memory Pools chart is a clear indication of object allocations While you are recording data you expect objects to be allocated Unless your hypothesis is that the objects will persist you should expect these objects to be removed at the end of the session A graph that does not drop to pre recording levels particularly after requesting one or more garbage collections warrants further investigation A large number of allocations may also prompt you to investigate the efficiency of the method you may be allocating more objects than necessary Starting an application server can cau
160. alling Context Tree 222 C Call Graph customizing 177 187 323 customizing fonts 324 hotspots 247 navigating by hiding and showing nodes 320 node types 297 pruning subtrees 319 threads and methods 249 viewing calling relationships 298 Call Graph tab 298 Call Traces customizing 177 187 Call Traces report 207 Call Traces view 121 174 allocated live instances 176 interacting with 178 opening the 175 understanding the 175 Call Tree isolating 228 pruning 228 threads and methods 249 Call Tree tab 302 calling context sites 330 Calling Context Tree building 222 filtering and encapsulation 225 how JProbe builds 223 pruning subtrees 231 calling methods navigating in Performance 322 in the Call Graph 322 subgraphs 321 catch blocks color coding in Coverage Source view 385 filtering out data 391 categories adding 432 creating and managing 432 deleting 437 editing 436 renaming 437 CCT See Calling Context Tree changing trigger execution order 429 chart exporting data to a file 443 Memory Pools 221 saving as animage 443 child navigator 247 classpath specifying for a Java application 34 clearing snapshots 449 collecting data on lines of code 255 on methods 222 using triggers 256 collection references obsolete 109 color interpreting in Coverage Source view 385 setting for Coverage Source view 392 command line licensing 468 communicating results 439 compiling your program 415 concurrent licens
161. alog box opens Running Remote Sessions 461 Starting and Viewing a Remote Session Attach to Running Session Probe Engines mx Select the engine or enter the Host Name and Port Number below the table Name Product User Memory azafar Memory llupas com sun enterprise glassfish bootstrap ASMain Memory Host Name IP Address localhost Port 52991 x OK l Cancel l Refresh 5 Enter the host name or IP address of the remote server and the port number that the Connection Manager is waiting on and click OK The Console connects to the running session on the server You can now view the session as you would a local session For details about viewing data in the JProbe Console see the other sections of this guide 462 JProbe User Guide 2 JProbe License Options This chapter provides information about the license server that allows you to run JProbe sessions Before you can use JProbe you must provide a valid license file JProbe is unlocked for purchase or evaluation depending on the license file you provide JProbe has four types of licensing models available node locked per seat concurrent and enterprise For more information review the JProbe licensing options on the Quest Web site at http Avww quest com jprobe licensing_info aspx This chapter contains the following sections Node Locked LICENSING iin titi iieen Gatien Widen ceded onder Sete iaaiia
162. alysis Views 283 Exploring the Summary View Exploring the Summary View The Summary view presents an outline of the snapshot data and identifies potential performance hotspots You can use this view for guidence on how to continue the performance analysis investigation This section contains the following information e Opening the Summary View on page 283 e Understanding the Summary View on page 285 e Interacting with the Summary View on page 292 Opening the Summary View The Summary view is the first view you see when opening a performance snapshot To open the Summary view e Inthe Snapshots Navigator double click the snapshot that you want to display The Summary view opens If the snapshot contains JDBC data the Summary view presents the snapshot s information in two tabs Methods and JDBC tabs The following illustration presents an example of a Summary view when JDBC data is recorded 284 JProbe User Guide welcome Summary blll Methods JDBC Connections The table contains connections created using one of the java sql DriverManager getConnection String url methods Each row represents a url parameter value not a connection object The Calls column displays how many times getConnection was called with each URL The time columns display the time spent in getConnection for each URL Database URL Calls CPU cml Elap Cmi rbyDB create true 1 e 1
163. amount of data you need to assess You will see different results depending upon the type of encapsulation that you run with your session After running a session and investigating hotspots you will identify methods that are working well In subsequent sessions you can encapsulate data on these methods so that you can focus on areas of interest Filtering Encapsulated Methods Performance filters in the JProbe settings tell JProbe which methods are considered encapsulated By setting filters you can include only the methods that you want to view in the Call Graph In addition by using the Encapsulated action Performance controls how the filtered methods should be treated For more information see Refining Data Collection Filters on page 422 Running a Performance Analysis 251 Investigating Your Results Encapsulation Types There are three levels to choose from to collect encapsulated data Full Grouped default and None Each setting provides a different level of abstraction for viewing method information Encapsulating data assists you in focusing on hotspots and data that is important to you To begin investigating the encapsulated options let s examine a call tree representation of methods captured in your application The parent methods and their encapsulated children method are displayed in a hierarchical view Root Y Main a S v1 y YY X Ht H2 v
164. anager Config 28 6 demos gamepack framework config GamePack 57 1 Class and Method Summaries The Method Summary breaks down the Class Summary by class For each method in a class you see whether or not the method was hit by your test case and the lines of code expressed as a percentage that were hit Reporting Coverage Analysis Results 401 Snapshot Browser Report a Classes Name Methods Hit demos camepack ejb user UserBesn 16 7 demos camepack framework serviet GlobalServiet 90 0 gdemos gamepack ejb game GameHandlerBean 143 demos gamepack framework session HttpoGameSessionManager 11 1 demos camepack fremework serviet LoainitServiet 75 0 demos camepack framework util XmiUtil 100 0 demos camepack framework GameGlobal 33 3 demos framework 2 JDBCDatabaseManacger 11 8 demos cemepack framework config GamePackController 714 demos gamepack framework config ManagerProperty 100 0 demos gamepack framework config GamePackGame 58 3 demos camepack eib came GameBesn 14 3 demos camepack framework user E BUserManacer 14 3 demos gamepack framework config GamePackManager 87 5 demos gamepack framework confic GameMode 71 4 demos qamepack fremework user E BUserManager Config 28 6 demos gamepack framework config GamePack 57 1 Per method summary for class demos gamepack ejb user UserBean Name Method Hit Lines Hit lt int gt false 0 0 addPoints int User false 0 0 getUser String false 0 0 addU
165. analysis e Data Recording Off e Public Method No Encapsulation Learning More About Performance Options 269 Analysis Options for Performance e No Encapsulation e Grouped Encapsulation e Full Encapsulation Encapsulation Policy An encapsulation policy determines how data for methods marked encapsulated is collected and how that data is attributed Encapsulation can be specified as a Recording Parameter when you are adding a trigger For more information see Triggers for Performance on page 271 After you finish running the session you will examine the resulting snapshot of data in the Methods view The encapsulation policy you selected affects how some nodes are displayed in the graph The selected policy is displayed in the status bar of the Methods view For more information see Exploring the Methods View on page 297 To help distinguish among the available policies consider the following call sequence A non encapsulated method methodA calls out to an encapsulated method method which in turn calls other encapsulated methods methody and methodZ methodz then calls another non encapsulated method methodB Policy Description Call Sequence Advantage Example Disadvantage None JProbe aggregates data The data for With this policy data for the non method collection is the fastest encapsulated calling methodx It produces a simplified method plus the methody and Call Graph but the data encapsulat
166. ance diners Philosopher run 9 2 demos performance diners PhilosopherPanel drawPhilosophers 0 0 demos performance diners Philosopher paint java awt Graphics 0 0 demos performance diners PhilosopherPanel stopRequest 0 0 demas performance diners PhilosopherPanel paint java awt Graphics 0 0 What Next To investigate a method in one of the tables double click its table row To investigate any method enter its name here classname methodname x Exploring Performance Analysis Views 285 Exploring the Summary View The following illustration presents an example of a Summary view when no JDBC data is recorded and both Elapsed Time and CPU Time are available welcome B Runtime Summary RE Summary v Ral 7 Performance Snapshot Taken Sep 28 2009 2 49 00 PM Timers Elapsed CPU Recorded Level Full Encapsulation Elapsed Execution Time of All Threads Combined A 27m Recorded Time Sm CPU Execution Time of All Threads Combined B 93ms CPU Utilization Rating 0 0 X Performance Hotspots Determine Hotspots Based On i O CPU Time Methods to Investigate Name Method Time of A demos performance diners Philosopher run 9 7 aaa a java lang Thread sleep long 0 3 java io PrintStream printin java lang String 0 0 java awt Graphics drawImage java awt Image int int java awt image ImageObserver 0 0 java awt Graphics drawString java lang String int int 0 0 Methods That
167. and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 e Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 Requesting a Module from Quest Support Modules are provided on request as a service to Quest Software customers with valid support contracts However Quest Software cannot guarantee that all application servers can be supported or that the module can be ready for you within your timeframe 68 JProbe User Guide If time is short you may want to try one of the other options to integrate your application server To request and install a module for your application server 1 Log a support case Visit SupportLink and select Case Management Create a case and request a module for your application server 2 When you receive the module JAR file save it in the JPROBE_HOME nodules directory 3 Launch JProbe The new application server is displayed in the list of available servers in the Configurations Navigator 4 Create a configuration as usual Note Additional modules may also be made available from the JProbe Community at hittp jprobe inside quest con Modifying an Application Server Startup Script You can also configure JProbe for your application server by modifying your application server s startup s
168. ands are e run Start Catalina with JProbe in the current window e run security Start in the current window with security manager start Start Catalina with JProbe in a separate window e start security Start in a separate window with security manager 80 JProbe User Guide Apache Tomcat v6 0 When creating your JProbe configuration you indicated whether you start Tomcat as a Windows service using tomcat exe or from a startup script You will run the JProbe generated startup script differently depending on what you indicated Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Start Tomcat as a Windows service tomcat exe Usage lt configuration_startup_script gt commands where commands are e run Default Start Catalina with JProbe in the current window e run security Start in the current window with security manager start Start Catalina with JProbe in a separate window e start security Start ina separate window with security manager Start Tomcat from a script If you start Tomcat from a script you can run the JProbe generated startup script in the same way that you run your Tomcat startup scrip
169. appears For more information see Allocations Detail Report on page 206 Exploring the Instance Detail View The Instance Detail view allows you to extensively analyze the instances of a class You can see the instance s reference and referrer trees as well as a class list of dominating instances This section contains the following information e Opening the Instance Detail View on page 166 e Understanding the Instance Detail View on page 167 e Interacting with the Instance Detail View on page 171 Opening the Instance Detail View You can drill down to more detail on the instances of a class from several different views The Instance Detail view is reachable from most Memory views To open the Instance Detail view 1 Select a row and click Instance Detail View t The Instance Detail view opens The instances for the selected class or method are displayed in the upper pane To open the Instance Detail view with a single instance 1 Select an instance in the Heap Graph 2 Do one of the following e Double click the instance e Click Instance Detail View The Instance Detail view opens with the selected instance in the instance list Exploring Memory Analysis Views 167 Exploring the Instance Detail View Understanding the Instance Detail View The Instance Detail view has two panes The upper pane displays a list of instances collected during data recording The bottom pane has three
170. apshot drop down list choose the snapshot with the modified code The Snapshot to Compare drop down list already displays the snapshot with the original code Click OK The Performance Difference view opens Use the method filter and column sorting to find the entry method for your application The cumulative time for the method is the overall performance of your application If the overall performance has improved your modifications have been successful Determine if the percentage reduction in time objects is enough or if you need to attempt further improvements If the overall performance is the same or has deteriorated investigate individual methods Note where you reduced time objects and where you increased them Determine if you can make further improvements or if you should revert to the original code Close the Performance Difference view Identifying Deadlocks Performance can report thread deadlocks in your Java SE and Java EE applications This section includes the following topics Understanding Deadlocks Finding Deadlocks Running a Performance Analysis 261 Investigating Your Results Understanding Deadlocks A deadlock is a situation in which a cycle of threads is blocked because they are waiting for access to a resource that is held by another thread in the cycle and so can never be freed The program can never terminate because the threads are blocked indefinitely With standalone applications you exp
171. apshot during the session click Take a Snapshot w 8 On the Runtime Summary view click the Heap Data tab 116 JProbe User Guide 10 11 QWelcome A Runtime summary Tawa vT kee ee configure Pools lonann fae Configure Pools Pools and GC Interval One Minute RS Classes Loaded 1174 Classes Unloaded 0 Memory Pools GC Data Heap Data Investigate by Heap Count Clickon chart segments or legend items to drildown on categories Categories Bava Base Libraries DpPrimitive Arrays Hava Ur Toolkits Bother Java SE azi Hother 50 60 Name Filter 3 625 6251 To Refresh the Runtime Data Press Class Recorded Count Heap Count Recorded Memory Heap Memory Dead Count Dead Memory Totals 71 21 862 2 896 1 646 072 12 485 490 488 java lang String o 4 347 0 104 328 210 5 040 n chart 4 4 310 312 257 624 462 23 200 J short a 1 841 o 127 000 1 16 it o 1 414 f 118 064 3 816 byte o 1 352 0 205 840 33 5 376 java lang Class a 1 331 o 567 960 o o java util HashMap entry o 1 017 o 24 408 198 4 752 This view is similar to the Instances view see Exploring the Instances View on page 138 except for the Refresh Runtime Data button and the information below the table To display the data that has been collected click Refresh Runtime Data E Refresh Runtime Data OK JProbe turns the data
172. ar call you hover over an edge between two nodes and a pop up dialog appears containing the method name and data about the call Alternatively the Method Detail view also displays the method names and all of the information is displayed in a summary area You can also view additional information about the individual nodes in the CCT table in the Context Calling Sites CCS For more information see Building a Calling Context Tree Summary The top left pane displays summary information about the method that you are viewing The Show Source 3 button allows you to view the code for the method Summary Item Description Current Method The name of the method Calls Number of calls to the method Exception Exits The number of times the method exited due to an uncaught exception rather than a normal return Catches The number of exceptions caught by the method Method Time The amount of time that it took for the method to execute Child Time The amount of time that it took for all children to execute Cumulative Time The combination of method time and child time CPU Method Time The CPU resources used by the method to execute CPU Child Time The CPU resources used by all children to execute 328 JProbe User Guide Summary Item CPU Cumulative Time Description The combination of CPU method time and child time Recursive Calls Number of recursive calls Recursive Time
173. arly in this view For example if your young generation is too small or objects are being moved to the old generation too soon this will show in the view The young generation will fill often and the objects will be tenured to the old generation so the gauges for the pool or pools in the young generation will fill and empty regularly Since short lived objects will have been moved to the old generation too soon you will see it regularly growing with the new objects then shrinking as it garbage collects them Over allocation of objects may also be visible as spikes of Used memory in one or more of the pools in the Memory Pools chart Reducing Over allocation of Objects Over allocation of objects can have a performance impact either by affecting GC overhead or if the objects are very expensive to create for example database connection objects If you suspect over allocation may be a problem JProbe can help find the source code that is causing the allocations so that you can correct it It is important to note that over allocation is not always a problem JVMs are tuned to optimize allocation particularly of short lived objects Some techniques used to reduce allocation such as object caching can actually have a negative effect on garbage collection so it is important to consult vendor documentation and to measure to see if your changes do improve performance As mentioned above over allocation may be visible in the Runtime Summary as
174. arning More About Coverage Results on page 393 To view the source code double click the method You may be prompted to locate the source code file If you have the source code a Source view opens with the code displayed By default missed statements are displayed in red text and missed conditions if tracked in gold Lines of code that were exercised by your test suite are in black Running a Coverage Analysis 365 Investigating Your Results text Lines with no data are in gray If you are filtering out catch blocks these are displayed in gray as well welcome amp Coverage Session fj Snapshot Browser Kitchen java Methods Kitchen lt init gt X Line Missed Source 4 public Kitchen 36 Yes setName 8 37 Yes setNumDirections 3 38 Yes setDescription You ve made it to the kitchen Grab a quick snack 39 Yes setDirections new String north east west 40 Yes setAssociatedNames new int 4 6 5 7 End of class Catch blocks not filtered Along the right margin is a navigation aid called the Coverage Bar The length of the Coverage Bar represents all the source code in the file not just the code currently displayed in the view The colored stripes on the bar represent the relative size and position of missed lines of code and missed conditions if available You can click a stripe to navigate to that block of code or use the arrows at the
175. ass e Method Class the methods that pass the set filter The Filter Methods field above the Method List displays the name of the selected method package or class Exploring Performance Analysis Views 317 Exploring the Methods View Task Prune a subtree Action select a node in the Method List right click and select Prune Selected Node Response The selected node and its subtree are removed from the graph All metrics are recalculated without the contribution of the pruned call tree For more information see Pruning Subtrees on page 319 Isolate a subtree Isolate Selected Node The selected node is anchored to the root node All nodes that do not belong to the method s subtree are removed All metrics are recalculated to include only the data collected on the methods in the isolated tree For more information see Isolating Subtrees on page 319 Re draw the graph using the select node as the focus method Redraw Graph Based On Selected Node Resets the graph clears the history and re draws it using this node as the focus method Show hide table columns Show Hide Columns Select the columns to show and click OK The selected columns are displayed The rest are hidden Show hide all columns displaying Elapsed Times Show All Elapsed Columns All columns displaying Elapsed Times are displayed hidden JProbe User Guide
176. ate the cost of the methods called by the Current method by reviewing the information in the Edges To Children area 5 You can navigate by using the navigation arrows and by double clicking any method The Method Detail changes to show the new selected method in the Method area The Parents and Children tables now reflect data in terms of this method Note To return to the last Current Method select the back arrow or click the forward arrows to navigate to the next method in our method viewing history 6 Evaluate the code for the method itself by clicking 3 Show Source The Source view opens with the source code containing your method Running a Performance Analysis 249 Investigating Your Results Next Steps Consider your alternatives for improving the speed of the algorithm contained in this method You may be able to modify the algorithm to run faster cache results for future use or change the way you call other methods Rerun the session with the modified code using the same JPL settings file and take a performance snapshot You can then use the Snapshot Difference view to assess your modifications For more information see Measuring Performance Improvements on page 258 By improving performance bottlenecks in the critical path you will achieve visible improvements in the runtime of your program If there is nothing you can do with this method repeat the procedures to find the next problem method Narrowing Your Invest
177. ated with the selected method opens Disabled if line data is not available for the selected method The following table lists the operations that you can perform from the context menu of a Call Graph edge Task Find the destination node Action select an edge in the Call Graph tab right click and select Go to Destination Node Response Selects the destination node that is the node to which the arrow points Note The node is highlighted in all three components Call Graph Call Tree and Method List JProbe User Guide Task Find the source node Action select an edge in the Call Graph tab right click and select Go to Source Node Response Selects the source node that is the node from which the arrow originates Note The node is highlighted in all three components Call Graph Call Tree and Method List Exploring the Call Tree The following table lists the operations that you can perform from the Call Tree context menu Task Expand a node to the next branch point Action select a node in the Call Tree tab right click and select Expand To Next Branch Point Response The selected node expands to the next branch point Move to the Method Detail view with a focus method Open Method Detail View The Method Detail view opens with the selected method as the focus method View the list of threads rep
178. available in a JProbe snapshot such as traces and garbage collection data Some imported dumps do not contain field names Also PHD and text dumps do not contain any information on roots so it is sometimes difficult to tell if an instance is actually pinned e Importing an IBM JVM Heap Dump on page 127 e Importing an HPROF Heap Dump on page 129 Importing an IBM JVM Heap Dump If you experience an out of memory exception or you want to take a snapshot of the IBM JVM outside of JProbe you can generate a heap dump You can then import it into JProbe and view it as a snapshot JProbe supports two different formats for importing heap dumps IBM Text dump txt or IBM Portable Heap Dump phd For more information see How to generate a heap dump from an IBM JVM The heap dump information that is saved to the snapshot contains classes and instances with their associated references and referrers Some objects in the heap dump have a path to a root and others have zero referrers but are not declared as roots Because neither text nor PHD format provides root information it is sometimes difficult to tell whether an instance is actually pinned or if it is ready to be garbage collected Therefore to provide a complete list of all objects in the heap dump objects that have paths to a root and non rooted objects are all included in the snapshot To import an IBM dump 1 Select File gt Import 2 Select one of the following options e
179. aw jph Recording Starts recording data In the Recording Parameter section you can select the level of data you want to record e Data Recording Off e Record Allocations For All Instances e Record Allocations and Stack Traces For All Instances e Record Allocations Stack Traces and Garbage Data For All Instances e Record Data as Specified in the filters Table Below Note When the trigger action is set to Recording the selected recording parameter is also displayed in the triggers table Request GC Requests a garbage collection Initial Recording for Memory You can select initial recording options for your Memory analysis in the Initial Recording tab The following Initial Recording at JVM Start options are available for a Memory analysis e For Java Applications e Data Recording Off default selection e Record Allocations For All Instances e Record Allocations and Stack Traces For All Instances e Record Allocations Stack Traces and Garbage Data For All Instances e Record Data as Specified in the Table Below e For all supported Application Servers e Data Recording Off default selection e Record Data as Specified in the Table Below Learning More About Memory Options 127 Importing Heap Dumps Importing Heap Dumps JProbe is a powerful tool for analyzing heap dumps You can import a heap dump and save it as amemory snapshot However imported heap dumps lack some of the data that is
180. be startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly 78 JProbe User Guide JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Usage lt configuration_startup_script gt run start geronimo_args where e run Start Geronimo with JProbe in the current window start Start Geronimo with JProbe in a separate window Note If you are running Geronimo 2 0 the port number is displayed in GERONIMO_OUT not in the command window Apache Geronimo v2 1 You run the JProbe generated startup script in the same way that you run your Geronimo startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Usage lt configuration_startup_script gt run start geronimo_args where e run Start Geronimo with JProbe in the current window e start Start Geronimo with JProbe in a separate window Note If you are running Geronimo 2 1 the port number is displayed in GERONIMO_OUT not in the
181. below the package category row and do one of the following e To select a subpackage or class click the cell browse button navigate to the package or class and click Open Set the Action to Exclude e To select a method click the browse button in the cell and navigate to the class containing the method When you select the class you can expand its list of 434 JProbe User Guide methods Locate and double click the desired method Set the Action to Exclude e Enter the package class or method no signature in the cell See the following subsection Syntax for Filters Set the Action to Exclude 6 To add existing folders or categories to this category choose the folder or category and click the Include button Note Order is important Filters lower in the table override filters defined higher in the table If you need to rearrange the filters click and hold the row button for the filter you want to move and drag it to its new location Right click a row button to insert or delete a row 7 Click Apply 8 Repeat to create additional categories 9 Click OK Syntax for Filters Filters are case sensitive and blank spaces are not allowed You can use an asterisk as a wild card it matches any character including the package separator If you specify the code in a shorthand or partial form JProbe automatically changes it to canonical form The change may not be exactly what you desire so be sure to verify it JProbe doe
182. by Dead Count Recorded Memory Traces JProbe collects data about object allocations and stack traces for the specified package or class In the JProbe Console data collected with this action is displayed in the Merged Allocations Points view and the Call Trace view when investigating by Recorded Count Memory Allocations Default JProbe collects the following data about object allocations for the specified package or class e Allocation time for each instance e Whether an instance was garbage collected e Recording boundary for instances that remain live when a snapshot is taken Exclude JProbe does not collect any data for the specified package or class Triggers for Memory Triggers provide a way to tie JProbe actions to your program s execution You can tell JProbe to perform an action when your program enters or exits a method of your choice With advanced trigger controls you can also specify whether to start executing the trigger actions at method entry or exit Triggers are useful for automated sessions 126 JProbe User Guide Triggers are defined in the same way for all the analysis tools however the trigger actions are somewhat different This section describes the trigger actions for a Memory analysis To learn how to create triggers see Setting Triggers on page 425 Trigger Action Snapshot Description Takes a snapshot The snapshot file name takes the form snapshot_ lt n gt _r
183. can change over time Configuring JProbe to Analyze Your Application 35 Specifying Your Java Application If you want JProbe to know about packages in the paths you specify select the check box beside the path JProbe can use this information to create default filters in the My Application Filters table for these packages For more information see Specifying Your Code on page 38 Tip If you create your settings on Windows and then run it on a UNIX platform the classpath will contain syntax errors because JProbe would have used semicolons between paths as required on Windows To avoid this problem you can create an environment variable on your UNIX platform that contains the classpath and then enter the name of the environment variable in the first cell Alternatively you can hand edit the settings file to change semicolons to colons To specify the classpath 1 Click the browse button beside the Classpath text box 2 To add the working directory to the classpath click Add Working Directory Classpath One Per Line Check the first column if you want the item in the second column to be used when proposing filters Note thatthe Classpath environment variable cannot be used to propose filters since its value is not known until the session is run Classpath One Per Line Program Files JProbe 8 1 o Add Working Directory Add Classpath Environment Va
184. can select the check box beside any paths that contain exploded code JProbe automatically displays these paths in the Elements table For more information see Classpath on page 34 To create application filters from elements 1 In the Elements table click an empty cell and do one of the following e Click the cell Browse button tJ navigate to an element and click Open e Double click the cell and enter the fully qualified path to the element 2 Repeat for any other elements 3 Click Create Filters 44 JProbe User Guide xn Oo oa A The My Application Filters table is updated with a list of filters based on your application elements By default all new filters are enabled and appended to the list of filters in the table Elements Enter the war ear and or jar files containing the code of your application Element IC Program Files ApplicationServers BE user _projects applications weblogic1 0 gamepack gamepack ear Insert Delete To use a filter select the row check box Set the filter Action to Include default or Exclude the code Repeat for any other proposed filters that you want to use Order is important Filters lower in the table override filters defined higher in the table If you need to rearrange the filters click and hold the row button located at the beginning of the row for the filter you want to move and drag it to its new location To delete a filter from the My Appli
185. cation Filters table 1 2 Select the row containing the filter you want to delete from the list Do one of the following e Right click the row button located at the beginning of the row and click Delete e Press Delete on the keyboard The selected filter is deleted from the My Application Filters table How JProbe Derives Filters from Elements Any file or directory can be selected as an element JProbe examines the elements files or directories and proposes filters for all java classes it finds You get a mixture of package class and method filters based on the depth of the package structure and classes existing in the package Users can select any file or directory and the filter proposal code will traverse all directories inside them Configuring JProbe to Analyze Your Application 45 Specifying a JProbe Analysis Type Element Code to include as an Application Filter WAR JAR EAR Includes code classes in ZIP or any other file e All directories type e The WAR JAR EAR ZIP file Directory path Includes code classes in the directory and its subdirectories When JProbe Does Not Propose Filters JProbe does not propose filters in the following circumstances e Ifthe file is not a valid ZIP archive file e If the file is not a valid class file e Ifthe elements directory does not contain any class file e Ifthe elements contain only code that is recognized as System Code or 3rd Party Frameworks categorie
186. ce Difference View Before opening the Snapshot Difference view you should load the performance snapshots you want to compare For details see Managing Snapshots on page 447 Caution Make sure that you selected valid snapshots you cannot compare a snapshot against itself or against a snapshot from a different application To open the Performance Difference view 1 Load the snapshots that you want to compare 2 In the Snapshots Navigator right click a snapshot and select Snapshot Differencing The Performance Difference dialog box opens The snapshot you selected is displayed in the Snapshot to Compare field Exploring Performance Analysis Views 339 Exploring the Performance Difference View t Performance Difference E3 Select Snapshots Baseline Snapshot A Snapshot to Compare Perf_Snapshot_3 3 Select another snapshot from the Baseline Snapshot list 4 Click OK The Performance Difference view opens R Pertormance Orterence 7G Fiter Methods ETE Baseine Perf _Snapshot_3 Other Perf Snapshot 4 Name Nam Cats Cumelstive Time Mathod Time Y Avg Cumulative Tine Avg Mathod Time Num Catches MumExcapt demos performance draers ChopStich relessel 7 r 3 a a lt ROOT gt RO0T gt 0 43 295 197 o o o o demos performance diners Phisogher gotHonMany Chopsbichslint s 0 o o o demos performance diners Phiosocher whichrhosooher int 1 005 0 e o o demon performance ners
187. centages The percentages above the table are calculated using the following formula Percentage Compared Baseline Baseline 100 If the resulting value is gt 1000 or lt 1000 the change is reported as huge Using Filters in the Memory Difference View If you want to specify filters you can use the Set Filters dialog You can re use existing filters or de select them so that they do not filter data You can also use the Filter Classes option above the table which is described under the Instance view see Filtering Data in the Instances View on page 141 To see how data is affected by Set Display Filters consider the following example of how the Memory Usage Change By percentage is calculated first with filters applied then without The totals in the snapshots are different For example the baseline snapshot with filters has a total of 50 instances while the total without filters is 100 instances Memory Filtered Data All Data Total in Baseline Snapshot 50 100 Total in Compared 40 90 Snapshot Usage Change Calculation 40 50 50 90 100 100 100 100 Usage Change Percentage 20 10 Therefore when comparing Memory Difference results be sure to take the filters setting into consideration Interacting with the Memory Difference View You interact with the Memory Difference view in much the same way as you interact with the Instances view The following tabl
188. ces Graph is a traditional call graph It displays a call graph of the program for call paths that allocated instances that are still live When the graph expands beyond the size of the pane you can use the Panner to focus on various areas of the graph or use the scrollbars to move around the graph You can also use the Scale slider below the graph to enlarge or shrink the graph When you click a node in the graph the corresponding line in the table lower pane is highlighted and vice versa The Instance Detail toolbar button then becomes available for further investigation of the instance The default status color scheme for the Traces Graph is Cumulative Count You can select an alternate scheme Method Count from the Color to Show list above the graph You can hover over the color scale at the bottom of the graph to see what they represent The lower pane displays a list of methods that have allocated live instances during a session The table collects data for the methods in the tree or graph into a single entry Data is displayed in the following columns Data Description Name For each row displays the names of the locations that created the objects Depending upon how the data is viewed the name represents method class or package data Cumulative Count For each row displays the number of live instances allocated by the location directly or indirectly by calling other methods Count For each row displays the numb
189. ces the list of threads in the lower pane so that only threads in that group are displayed You can return to the full list by clicking the Show All button The total number of threads is displayed beneath the graph along with the time the graph was last updated Historical data is not maintained for the graph The lower pane displays a list of all known threads grouped into categories This list is sorted in ascending order threads followed by categories When you click a thread in the list details for the selected thread are displayed on the right pane including e Name of the thread e Thread state e The object upon which the thread is blocked when applicable e The thread that owns the object that the thread is blocked when applicable e Stack trace of the thread e List of object monitors locked by the thread e List of synchronizers locked by the thread You can reduce the list by clicking a bar in the graph or by typing a filter into the Filter Threads field The filter is a simple string filter that matches against thread names not category names Categories can be expanded by selecting them right clicking and selecting Expand Category Threads can be collapsed back into categories by selecting them right clicking and selecting Collapse Category Thread details are only updated on demand You can update the details for the currently selected thread by clicking the Update Thread Details button or by clicking a thread
190. ceseces 107 089 0 oa 303 execution Searching is not available in the tree but you can sort the columns to find the methods that are using the largest amounts of time A time value of zero 0 can be caused by three things e Rounding occurs when the timing unit or precision does not allow for enough decimal places to display the precise value Very small amounts of time can be rounded down to zero Rounding is a display issue only calculations are not affected Changing the Time Unit may prevent rounding see Customizing the Call Graph on page 323 e Insufficient timing resolution means that the fraction of time spent in the method is less than the time measurable by JProbe e A method that uses a small amount of time 304 JProbe User Guide The name of each method in the Call Tree is prefixed by an icon which provides details about that method as described in the following table Icon Type Root nodes optional Thread nodes Grouped thread nodes Normal simple nodes Encapsulated grouped nodes Note The methods names match their parent s names H DM OD Aggregate nodes m JDBC nodes Note The status bar also displays the most common icons and short descriptions of what they When you use the prune or isolate functions in the Call Graph the corresponding nodes are removed from the Call Tree However if you use those fun
191. ch as a WAR EAR or JAR file click the cell browse button and navigate to the element When you select the element you can expand its contents Locate and double click the desired code e Enter the package class or method no signature in the cell For more information see Syntax for Filters on page 41 The checkmark at the beginning of the row indicates that the filter is enabled Tip To enable disable all defined filters at once you can select clear the check box beside the Data Collection Filters table title 3 To set the data collection action that should occur for the specified code click the Action cell and select an action For more information about actions see e Filters for Memory on page 124 e Filters for Performance on page 267 e Filters for Coverage on page 369 Tip You can interact with the Action cell only after you specify a method 4 Click Save Syntax for Filters Category filters use the same syntax as other JProbe application filters For more information see Syntax for Filters on page 41 Using My Code Packages to Add Filters You can add data collection filters to the current configuration based on the application filters that you specified in other configurations Recall that application filters are set during the configuration wizard on the Specify Your Code screen To add filters from previously defined configurations 1 Click Add My Code The list contains your
192. cifying Advanced JProbe Options on page 46 In this screen you have the following options Continue the configuration process by reviewing the settings Click Next Continue on step 10 Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 10 Optional Review the settings in the configuration summary screen For more information see Reviewing the Configuration Summary on page 49 The configuration is now complete In this screen you have the following options Edit any of the configuration settings if necessary Click Edit beside the setting that needs further updates The wizard opens the screen corresponding to the selected setting then guides you through the appropriate configuration steps Caution You may need to define some additional settings in the subsequent wizard screens for example filters and elements Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the
193. collected is saved in a temporary snapshot 9 To end the session click Detach From Session F on the JProbe Console toolbar then exit your program as you normally would The Summary view opens with the most recent performance snapshot displayed Saving Baseline Snapshots You can save the snapshots you take during the exploratory session so that you can use them later on as a baseline A baseline will help you to quantify improvements to yourself your team members quality assurance product management marketing and other stakeholders For more information see Measuring Performance Improvements on page 258 Note You may need to demonstrate improvements against an externally set benchmark JProbe is a pre production performance analysis tool not a benchmarking tool your JVM may have performance characteristics that are very different when run with and without the profiling interface In this case you should use JProbe to set a baseline find performance bottlenecks fix the code and verify the performance improvement against the baseline To measure against the benchmark rerun the fixed code without JProbe in the benchmarking environment and collect timing data The JVM you select and JVM options you set can influence how your program runs your program will likely run differently under JVMs from Sun IBM or BEA because vendors optimize data and implement different garbage collection strategies in diverse ways When runnin
194. columns to show and click OK The selected columns are displayed The rest are hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for the metric The table is sorted To reverse the sort order click the heading again Export the table data to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers Create a report for this view Click Export Data to HTML a The Create Performance Difference Report dialog box appears 342 JProbe User Guide Setting Preferences for Performance Views You set data display options in the Edit Options dialog box Option Show Percentage Time Values in Methods View Trees Description Defines if the results are displayed in the Methods view as percentages or as actual time values Time Unit Defines the unit for measuring time intervals displayed in Performance views The f
195. command window Starting Your Application with JProbe Attached 79 Runtime Notes for Application Servers Apache Tomcat v5 5 When creating your JProbe configuration you indicated whether you start Tomcat as a Windows service using tomcat exe or from a startup script You will run the JProbe generated startup script differently depending on that what you indicated Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Start Tomcat as a Windows service tomcat exe Usage lt configuration_startup_script gt commands where commands are e run Default Start Catalina with JProbe in the current window e run security Start in the current window with security manager start Start Catalina with JProbe in a separate window e start security Start in a separate window with security manager Start Tomcat from a script If you start Tomcat from a script you can run the JProbe generated startup script in the same way that you run your Tomcat startup script JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Usage lt configuration_startup_script gt commands where comm
196. cquire and release the monitor associated with a synchronized block is normally attributed to the start and end of the block respectively There is also however an implicit exception handler catch block generated to release the monitor in the event of an exception The exception handler bytecode may be attributed to either the start or end of the block JProbe identifies the affected lines in the synchronized block as a catch block 396 JProbe User Guide Nonintuitive JProbe Results Data is assigned inconsistently to one or more of the lines that make up a statement Explanation of Compiler Behavior When source code for a statement is split over multiple lines the compiler may attribute the bytecode for the statement to one or more of the lines of source code For example consider the following two statements for i 0 someLongCondition i i String str prefix tor suffix The compiler may attribute code to all three lines in the for statement but only to the first line in the String concatenation Uninitialized variable declarations are reported as missed rather than unexecutable When creating the line table for a method a few compilers assume that the start of the bytecode for the method can be attributed to the first line If this first line is a statement with no executable code such as an uninitialized variable declaration int i then no bytecode is created for the line Wor
197. cript manually If you feel comfortable making changes to your application server s script you can make a copy of it and replace the main call to java with a call to jplauncher instead Caution When modifying your application server script make sure that it is not making an external call to another script that starts the application server This will cause the configuration to fail Before you begin you need a JProbe settings file To create a simple settings JPL file with only JProbe options and no application specific information 1 Right click in the Configurations Navigator and select New Configuration The New Configuration Wizard dialog box appears 2 Click Settings File gt JProbe Settings File 3 Type in a name for this configuration in the Configuration Name text box and click Next Configuring JProbe to Analyze Your Application 69 What Can I Do if My Application Server Is Not Supported 4 Optional Specify a supported JVM plus any Java options you normally use and click Next For more information see Specifying a Java Virtual Machine on page 36 5 Specify application filters enable JProbe to create a category containing the code you are interested in set to include and click Next For more information see Specifying Your Code on page 38 6 Select the type of analysis you want to run For more information see Specifying a JProbe Analysis Type on page 45 In this screen you have
198. ct Size All Columns to Fit The affected columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for the metric The table is sorted To reverse the sort order click the heading again Export table data to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Create a report for this view Click Export Data to HTML E The Create Dominating Instances Report dialog box appears For more information see Dominating Instances Report on page 205 Exploring Memory Analysis Views 155 Exploring the Pinned Instances View Exploring the Pinned Instances View This section contains the following information e Opening the Pinned Instances View on page 155 e Understanding the Pinned Instances View on page 155 e Interacting with the Pinned Instances View on page 157 Opening the Pinned Instances View The Pinned Instances view can be reached only from the Instances view To open the Pinned Instances view 1 Select a class in the Instances view 2 Click Pinned Instances H or Right click the selected class and select the What
199. ction Response Set the metric by Select one of the options The currently displayed which nodes are colored available in the Color to Show list as necessary e Hot Paths Elapsed Time CPU Time to display the call tree that has the largest impact overall on program execution time critical path e Hot Methods Elapsed Time CPU Time to display individual methods that have the largest impact overall on program execution time Hot Methods Number of Calls to find excessive method calls Note The Color to Show list is by default initialized to Hot Methods Elapsed Time CPU Time set of nodes are colored according to the metric you selected and the color scale tooltips now reflect values for this metric In addition node tooltips reflect values for the selected metric In the Method List the heading that corresponds to the selected metric is highlighted in a bright color If the column was previously hidden it is automatically displayed and highlighted Exploring the Call Graph You can explore a graph by scrolling or using the graph panner You can see more or less of a graph by zooming in or out on the graph or by changing the physical size of the graph pane The following table describes how you can explore the Call Graph Task Increase the area of the graph Action Drag the bottom edge down Response The graph resizes to fit the larger area Explorin
200. ctions in the Call Tree the Call Graph is not changed although the underlying data for the corresponding nodes does change The only exception is if the corresponding node in the graph represents only one call in that case it is removed from the graph Exploring Performance Analysis Views 305 Exploring the Methods View Method List The Method List is defined as a fixed two level tree table The top level represents a graph node and the second level represents a CCT node A selection in the Method List can be mapped to the Call Graph and Call Tree and vice versa Note When the Methods view is first opened the Call Graph displays five nodes To display additional nodes use the Show More Nodes toolbar button see section Exploring the Methods View Toolbar on page 306 Each row in the list represents a method called by your program during its execution The data set may have fewer methods if you edited the data set by pruning subtrees or by isolating a subtree The Filter Methods field enables you to filter data based on the package or class of a selected method If a method in the list is not included in the Call Graph you can add it The name of each row in the Methods List is prefixed by an icon which indicates information about the method as described in the Call Tree section Interacting with the Methods View Several operations are available from the context menu of a selected item in any of the three components of t
201. ctly Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Usage lt configuration_startup_script gt options where options can be among others e user e passwordfile The startup script backs up domain xml and copies lt configuration_startup_script_name gt domain xml to domain xml After starting the domain JProbe restores the domain xmi file to its original state and deletes the backup Caution Do not end this script using Ctrl C It leaves your application server in an unstable state If the server is already in an unstable state restore the original domain xml from the backup copy Open Source Java Service Wrapper v3 3 The Java Service Wrapper is an application used to install e on MS Windows systems a Java Application or a Java Application Server as a Windows Service 90 JProbe User Guide e on Unix systems a Java Application as a daemon process This module can be used to start different Application Servers and Java Applications configured with the Java Service Wrapper Tanuki Software The JProbe configuration module requires two attributes to be set to integrate the Java Service Wrapper with J
202. d again in the child time To resolve this issue we modify our definition of cumulative time slightly to take recursion into account The cumulative time for a method is its self time plus its child time excluding any recursive calls Now the cumulative time is 15 30 7 52 10 Running a Performance Analysis This chapter describes how to run an analysis session and investigate your application s performance This chapter includes the following sections Setting Up JProbe for a Performance Analysis Session ssessessessesseseeseeetesneetesneeteeneesteneeass 236 Setting ihe Recording Level i aciaiids wid iaia mated aanedns 236 RUNNING AN Analysis SESSO iiss aida we anaian wnanddannasndddniddnnnneioaans 237 investigating Your RESUS seca aa eaen ORARE 243 236 JProbe User Guide Setting Up JProbe for a Performance Analysis Session Before you can run a Performance session you create basic session settings for your application according to the instructions in Configuring JProbe to Analyze Your Application on page 23 Later you can analyze individual methods with line level precision or measure the time the CPU actually spends on each thread For more information see Narrowing Your Investigation on page 249 When you run the JProbe analysis session your program launches or your application server starts For programs that require input to run you now exercise your program By focusing on areas with reported
203. d be eliminated from your profiling results In the following graph image x4 represents class loading activity A 8 4 2 X1 4 3 2 X4 1 1 1 CCT By pruning the x4 node the following results appear X4 node pruned Notice that the cumulative time for the remaining nodes has changed since the time for X4 is no longer included Isolate and Prune Operations Prune and Isolate operations affect the current CCT not the original one In the proceeding example method A was isolated and then additional methods were pruned from its isolated subtree not from the original CCT In this example an Isolate and then a Prune was performed on the result of the Isolate not on the original CCT If for any reason you do not want to save the results use the Back button 232 JProbe User Guide Prune and Isolate Subtree are available from the Call Graph any of the lists of methods beneath it or the toolbar Pruning or Isolating a method that calls multiple nodes affects all the nodes For example if you pruned x1 on the original CCT both the x1 node and x4 node are removed Pruning X1 By using Prune and Isolate commands we can view or exclude data from particular threads To look at a single thread you can isolate on its thread node To exclude particular threads for example application server threads you can prune them Recursive Methods Recursion occurs when a method calls itself directly or indirectly In
204. d set the action to include For more information about stack traces see Setting the Recording Level on page 110 and Analysis Options for Memory on page 124 If you have set the recording level to collect traces and still see the No Stack Trace Data message it is possible that the JVM is not making traces available to the JProbe Analysis Engine Disabling the JVM s runtime or just in time compiler may help for example Xint on some JVMs Operation Not Valid for Categories The tree graph or table may group multiple nodes by category and display just the category name to help you focus on the methods that matter to you Certain operations cannot be performed on categories You can drill down into these grouped items by selecting them in the table or tree and clicking Replace Category with More Detail E on the toolbar For more information about how to manage categories see Creating and Managing Categories on page 432 Strategies for Performance Analysis This chapter describes strategies for using the JProbe Performance analysis tool to uncover performance issues in your application data Tip Use JProbe Memory first to investigate how your program uses memory and to streamline garbage collections This chapter contains the following sections Strategies for Achieving Program Efficiency cesessesestessesssesesteseeseesteetesneetesnseteneeateneease 218 Understanding the Memory Pools Chart
205. d your copy of the domain xm file When you run the JProbe generated startup script you can specify parameters that you usually pass to the start domain command of asadmin bat Do not however specify 88 JProbe User Guide the domain name on the command line because it is set in the configuration If you do specify a domain name the application server will not run correctly Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Usage lt configuration_startup_script gt options where options can be among others e user e passwordfile The startup script backs up domain xml and copies lt configuration_startup_script_name gt domain xml to domain xml After starting the domain JProbe restores the domain xmi file to its original state and deletes the backup Caution Do not end this script using Ctrl C It leaves your application server in an unstable state If the server is already in an unstable state restore the original domain xml from the backup copy Open Source GlassFish v3 The GlassFish v3 requires the com jprobe filter string to be added to the org osgi framework bootdelegation property in the
206. dding a node to the tree for the method if necessary When a method is exited it takes the difference between the current time and the start time and adds this to the node s cumulative time Only cumulative time is tracked as the program runs method time is calculated later The JProbe Call Graph The Call Graph is a tool for viewing the CCT to show the calling relationships among methods in your program The Call Graph can be thought of as a set of superimposed call trees For example if two methods call method A then method A appears as a single node with two parents If you remove the call tree for one parent called pruning method A remains because it has another parent The method list located below the graph contains a list of methods with data for all the metrics In the Call Graph times and call counts for the edges are displayed as a tool tip but the data is not as rich as data found in the CCT Strategies for Performance Analysis 225 Understanding Performance Data v x1 sal c42 x2 3 2 xs 2 2 LJ X4 1 1 Call Graph The JProbe Call Tree The Call Tree presents the same data as the Call Graph in a traditional more detailed tree view While the graph displays only one node per method the tree view has a node for every distinct call path When you use the prune or isolate functions in the Call Graph the corresponding nodes are removed from the Call Tree However if you use those functions in t
207. deeper than you expect For example Method C line 300 Widget w new Widget 214 JProbe User Guide It is possible that JProbe will report the allocation as being in the Widget constructor Widget lt init gt instead of method C but method C will likely be in the trace calling the Widget constructor Memory Session Runtime Tips The following tips will help you improve session runtime performance regardless of the type of analysis session you are running Try different supported JVMs The session runtime performance depends in part on the speed at which your Java program runs Modern JVMs such as Java 6 generally improve program performance For more information see the JVM documentation and the Installing JProbe guide Exit your program before investigating snapshots If you open a snapshot in one of the investigation tools while your application continues to run you may find that the JProbe Console responds more slowly than desired To improve JProbe Console performance exit your program before opening a snapshot Alternatively you can disconnect from the session if you want your program to continue running Choose not to monitor the session Monitoring the session requires an extra JVM be running the JVM running the JProbe Console You can run a JProbe session without the JProbe Console Instead of interacting with the session through the JProbe Console you use triggers to collect data and take snapshots Af
208. dentify what objects are live using the Heap Data tab on the Runtime Summary view then decide whether you want to proceed by taking a snapshot to answer the question why is this live and include a heap dump or by beginning recording to answer the question where do these come from This helps you decide whether you need to use some filters for your recorded data In the Runtime Summary view click the Data Recording icon 2 The Set Recording Level dialog box appears Note The Set Recording Level dialog box is always initialized to the value you have last set To see the settings in effect in the JProbe Analysis Engine check the Session History for details see Viewing the Session History on page 445 Running a Memory Analysis 115 Finding Memory Leaks Set Recording Level Ob CO Record Allocations For All Instances CO Record Allocations and Stack Traces For All Instances Record Allocations Stack Traces and Garbage Data For All Instances CO Record Data as Specified in the Table Below Pattern L HEnaes J _ 6 If you want to change the Recording Level select one of the available options for details see Setting the Recording Level on page 110 and click OK If you were collecting data from startup this data is written to a snapshot If you are happy with the current recording level and want to exit without writing a snapshot click Cancel 7 To take a sn
209. der 1 Select Tools gt Manage Categories 2 Select a category right click and select Rename Note A Category created automatically by JProbe should not be renamed or at least a folder with the original name should be preserved otherwise JProbe creates it again 3 Specify the new name and click OK 4 Click OK Deleting Categories or Folders To delete a category or folder 1 Select Tools gt Manage Categories 2 Select a category right click and select Delete 3 Click OK 438 JProbe User Guide 23 Communicating Results JProbe offers many ways to communicate your results to management colleagues and Quest Support in case you encounter any issues This chapter presents how to create reports print to PDF save data to a CSV file and save charts as images using JProbe It also presents how you can view the analysis logs and the session history and how to send a report bundle to Quest Support The following notes highlight the differences among these options e Thecontent and layout of HTML reports is predefined HTML reports are created using valid XHTML 1 0 e Printing to PDF is WYSIWYG that is the output closely matches the current view You can control what data is included and the order of that data by setting up the view before printing e CSV comma separated values files contain raw data in a form suitable for import into other tools like spreadsheets You can control what data is included and the or
210. der of that data by setting up the table before exporting to CSV e Image files enable you to save and print out the charts You can set up the chart the way you want it before saving it as an image This chapter contains the following sections Generating HIME REPO 4c085 sana dnsadkn inl Aswan a aa aa E 440 Panting Data to PDF PG sseitiscnancdissaansdaannandduadnantddianddaadnanndan 441 Exporting Table Data to CSV File cessessessesssssesssssesnseeseneesteneeatensesteaseeaesnseateneeatenseatenseans 442 Exporting Runtime Chart Data to a CSV File ou esessessssssesstecnesssesssenseessessvessesnsecnsessersvenens 443 EEA ae atl TMS E E devennavetveds den deceadheiad E 443 Viewing the Execution NG oo as oe ee etc eceeceeecttns asec vdearatne ures ineiemapne ere 444 Viewing thie Session FISY i anini aaa aai 445 Generdmg a Suppo BU dl Eon ra a aa aa 446 440 JProbe User Guide Generating HTML Reports You can generate HTML reports based on many of the JProbe views For example you can generate a report on the Instances view or the Snapshot Browser view The options in the Create Report dialog box differ depending on the analysis tool and the report being generated JProbe creates a report with a header that includes the report title the version of JProbe and the snapshot name location and timestamp The content of HTML reports is predefined however JProbe respects the table sort order view filters and type of units that
211. e overview 469 using 469 conditions details 386 configuration copying a 59 creating anew 23 creating from a settings file 61 definition 29 deleting a 61 editing a 60 managing 59 renaming a settings file 63 resolving errors 61 saving 49 50 saving and integrating 55 Configurations Navigator overview 419 configure JProbe Java EE application 24 Java SE application 27 to analyze an application 23 configuring Memory Pools 456 Connection Manager 417 Console 417 console does not connect 95 contacting Quest 21 Context Calling Tree collecting data 223 recursive nodes 232 view in the Performance Call Graph 224 copying a configuration 59 Coverage analysis options 368 analysis strategies 355 assessing an existing test suite 356 bar charts hiding 391 catch blocks filtering out 391 code changes 357 color of results setting 392 concepts 393 creating a test suite 356 determining 356 filters 369 Index 475 general settings 368 hits in Coverage Session 375 investigating results 363 learning more about results 393 misses in Snapshot Browser 378 in Source view 385 modifying a test suite 356 percentages as bar charts displaying 391 recording controls 372 reports 399 running an analysis 359 running test cases 361 setting a basic analysis 360 Snapshot Browser 378 Source view 384 strategies 355 triggers 371 troubleshooting tips 410 views setting preferences for 390 Coverage reports defining f
212. e 7th time the method executes fire the 9th time the method executes fire the 11th time the method executes the trigger becomes inactive after it has fired four times or when the application ends whichever comes first Example Time based trigger control For example let s say that you want to wait ten minutes after the application starts and then have the trigger fire when the method is called Let s also assume that for the next twenty minutes you want to wait at least two minutes fire the trigger when the method is called then repeat To achieve this result you would choose Customize when the trigger action is performed and set the values as n 10 m 2 and x 20 The trigger would fire as follows wait 10 minutes from program start then fire the first time the method executes wait 2 minutes after the method call then fire when the method executes again repeat the wait method call fire cycle until 20 minutes have elapsed or the application ends whichever comes first the trigger becomes inactive after 20 minutes or when the application ends whichever comes first The procedure for adding a trigger control is the same for all the analysis tools To add a trigger control 1 Choose a trigger and click Advanced Control The name and type of trigger is displayed at the top of the dialog box Choose whether the trigger is invocation based or time based e To execute the trigger based on the number of times the method is
213. e Attached This chapter explains how to run your application under JProbe and connect a JProbe Console to the JProbe Analysis Engine It also presents how to attach a JProbe Console to an application started without JProbe This chapter contains the following sections Running a Java Application with JProbe 0 scsssssssssssssssnseessssnsetsassnsetsassassnsetsansnsetsatsnseteanenees 72 Running a Java EE Application with UProbe ou sssessssesssssessssnseteassnsetsassnssnsetsanenseteatsnseteananees 73 Runtime Notes for Application S rvers c ccsssssecsescseseetesnsetesneeeeneesneneestesseetesneeneeneeaneneeanenses 75 What to Do When the Console Does Not Connect eecessessessessessesseestessesneetesnsettsneeneneeateneeans 95 Connecting a JProbe Console to an Application Started Without JProbe cseesestessesesteeseees 96 72 JProbe User Guide Running a Java Application with JProbe When you integrated your configuration JProbe created a startup script To start your application under JProbe you simply run this startup script For more information see Integrating Your Application on page 51 If you are running a remote session you need to have JProbe installed on both computers Your license provides for this double installation To run your application with JProbe 1 On the computer where you run your application run the JProbe generated startup script from the command line or from the JProbe Console s
214. e Summary Heap Data Tab The Heap Data tab is similar to the Instances view see Exploring the Instances View on page 138 It contains a horizontal bar chart and a table with information on the instances allocated objects in the Java heap This includes the number of instances of each class in the heap at any given time and the memory used by those instances The top row in the table contains totals for each of the columns This table is not populated with heap values unless you have clicked the Refresh Runtime Data button Note The Heap Data tab is not available if you have connected to a JVM dynamically using Attach to Session gt JMX or jpattach Interpreting Heap Data The horizontal bar chart illustrates the percentage of instances in each of the categories that are listed in the legend to the right of the chart Hovering over a segment of the chart displays a popup showing the count or memory data for that category Each row in the table represents a class that has instances in the heap or has had all of its instances garbage collected The Count columns tell you how many instances of a class currently exist in the heap while the Memory columns show how much memory those instances consume The Dead Count and Dead Memory columns show values for garbage collected instances that were allocated since recording started Filtering Heap Data The Name Filter field allows you to reduce the set of displayed instances to only those
215. e VieW ssssssssssteseestsstesessnsseenseneeatententsaeeateateateatenteeteanestenteaneneennes 200 Setting Preferences for Memory Views eessessesssecseestecstessecstesseessecssecstesueeatecueeasecseeaueestecueessecseesseesesseesueestesneenseets 202 Reporting Memory Analysis R SuItS scsssscssssssssssersseseserseeeserseeseserseeeserseeseserseeaseesseeaseesseeaeenssneaseesseeas 203 Setting Report Options as anscae cakes EE TT vist tananneiaene natant 204 Runtime Summary Report vss eisai etal CR a ee a aes ee i ee 204 Instances Reporty aininn paa apaa a pa aR E aa AEA EAE E EAA EE Aan TERAS Daa tan eaa ERA diane 204 Dominating Instances Report siini E AEE EEE E EA E 205 Pinned Instances Report screener Aarseese heen erated ENEA AARE 205 Allocations Report raens nin a ni ei an cae ec dia ladies 206 Allocations Detail Report izsita ei dis ae eee ee cate 206 instance Deta RO OTT er a ara Er a aa aa a r ae e rA a EE A EE ER mie 207 Gall Traces Reports sean sissies a A a nt 207 Merged Allocation Points ROM 2 csncsicvr avarice aneeiiit aliens A a 208 Heap Graph Report d iere an aaaea arsi dni n a A ES E EE EAE EN NE EN 208 Memory Snapshot Difference Report sssesessssssssssrsttetetetseseeesssssssesesurtrtttttetetereesesssssssnsusnrnttnttteteteeesisnssssannnttttteteteeeeeete 208 Troubleshooting Memory Analysis Results scssssssssssesssseesesseeeeensneeeenrsneeeenrsteeesnnaneesennsneetennsnennaneenes 211 Memory Troublesh
216. e categories that are listed in the legend to the right of the chart Hovering over a segment of the chart displays a popup showing the memory used by that category Selecting a bar from the chart redraws the Pinned Instances view using the instances from the category chosen as the starting set The lower pane Pinned Instances by Type contains a table which shows the instances pinned by the starting set of instances listed in the upper pane For example if you go into the view with 30 HashMaps instances as the starting set you may see that they pin HashMap Entry HashMap Entry and often Strings The values displayed in the Recorded Count Heap Count Recorded Memory Heap Memory and Keep Alive Size columns in the lower pane are for the number of HashMap Entry HashMap Entry String instances that are pinned The middle pane summarizes the largest rows based on the Keep Alive Size value Exploring Memory Analysis Views 157 Exploring the Pinned Instances View The status bar displays the most common icons and short descriptions of what they represent In the Pinned Instances view you can navigate towards the leaves of the dominator tree The starting set shown in the upper pane pins a number of instances those instances are grouped based on categories in the middle pane and shown in detail in the lower pane You can drill down in a category by clicking either a section in the chart or the corresponding category
217. e covers the critical areas of your program It also helps to Strategies for Coverage Analysis 357 Keeping Up with Development reduce redundancy and makes the most of your testing effort You can add additional tests and work toward overall coverage of your application as the product matures Keeping Up with Development One of the pitfalls of parallel testing is that the code and the interface can change from day to day making some of your test cases out of date before you can run them By using JProbe you can be confident your test suite keeps pace with programming changes If you run a Coverage analysis daily and are familiar with results JProbe can help you to identify changes to the code The overall coverage percentage changes if new lines of code are added or removed New lines of code are identified as missed in your results 358 JProbe User Guide 16 Running a Coverage Analysis This chapter presents a basic analysis using default Coverage settings This chapter contains the following sections Setting Up JProbe for a Basic Coverage Analysis ssssscssessesseseeseestesseetesneetesneetenseateneeane 360 Running Test Gases wiht J PIOB iiini 361 relent ney VOUT ROSE fes thei eschew tacaane dixie ndae as enaermaeains 363 360 JProbe User Guide Setting Up JProbe for a Basic Coverage Analysis Note The following procedure assumes that you have already created a JProbe configuration If you have not configured JPr
218. e snapshot navigator as snapshot Session Disconnected The Coverage Session view shows data in terms of the number of methods and lines of code that were hit by your test case For more information see Exploring the Coverage Session on page 374 5 Work through your test case as usual 6 End your application as usual JProbe generates a Coverage snapshot that contains data on the test case 7 Repeat for each test case in your test suite 8 To make it easier to identify your snapshots you may want to rename them e To rename and save the snapshot right click the snapshot and select Save Snapshot As Save the snapshot with a meaningful name e To change the label only right click the snapshot and select Change Label Investigating Your Results After you have run your test cases you are ready to investigate your results In the Snapshot Browser JProbe displays results in terms of misses A miss means that the code was not executed during your test cases The screen captures in the following procedure are taken from the Adventure tutorial They are provided as an example only 364 JProbe User Guide To investigate your results 1 2 In the Snapshots Navigator double click a snapshot original or merged In the Snapshot Browser expand the package tree in the Application Tree pane top pane In the Missed Methods column select a class that contains the highest percentage of missed methods If you have
219. e startup script uses either a standard integration or a hard integration In a standard integration you will find that in one or more places where you would normally see a call to java there is instead a call to jplauncher The jplauncher is the JProbe tool that does the work of collecting data on your application For more information on jplauncher see the JProbe Reference Guide In a hard integration the startup script passes JProbe settings to Java as a Java option The settings file that was created from the configuration is not used directly Rather JProbe uses the jplauncher command jp_export_jpl1 command to create a JPL file that can be used directly by the JProbe Analysis Engine To execute the JProbe generated startup script 1 Inthe Integration Wizard dialog box on the Run Configuration tab review the startup script settings 54 JProbe User Guide Integration Wizard Run Configuration Run GamePack_Memory_WebLogic10 Path of script to start a java application or server with JProbe C Program Files ApplicationServers BEA_Systems user_projects domains weblogic10 gamepackistartWebLogicwithIJProbe cmd as Script Parameters ex C Program Files JProbe startwithJProbe bat Working Directory C Program Files ApplicationServers BEA_Systems user_projects domains weblagic 1 0 gamepack V Close syscon window on exit Required Field Run _ _ Run amp Connect cancel H
220. e summarizes the tasks that you can perform Exploring Memory Analysis Views 201 Exploring the Memory Difference View Task Filter the table Action See Filtering Data in the Instances View on page 141 Response The classes in the table match the filter criteria that you entered Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted Export the table to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 202 JProbe User Gu
221. ear all e To save some unsaved snapshots select the check boxes beside the snapshots and click OK Save the snapshots as usual e To save all unsaved snapshots click Select All and click OK Save the snapshots as usual Managing Snapshots 451 Managing Snapshot Transfers e To delete all unsaved snapshots click Clear All and click OK To return to JProbe click Cancel When the last snapshot is saved JProbe closes Any remaining temporary snapshots are deleted from the snapshot directory Managing Snapshot Transfers Temporary snapshots are sent from the JProbe Analysis Engine to the JProbe Connection Manager and then to the JProbe Console You can find out the status of the transfer To view the status of snapshot transfers 1 Choose Session gt Snapshot Transfer Status 2 To stop downloading the current snapshot click Cancel 3 To avoid downloading pending snapshots select the check boxes for the snapshots and click Remove Checked Selections 452 JProbe User Guide 25 Setting Preferences This chapter presents how to set preferences that affect how JProbe operates and displays data For preference settings specific to an analysis tool see e Setting Preferences for Memory Views on page 202 e Setting Preferences for Performance Views on page 342 e Setting Preferences for Coverage Views on page 390 This chapter contains the following sections Controlling How JProbe Operates
222. eases in the heap usage at certain points a method might be allocating a large number of objects The Call Traces view offers you a different way of looking at object allocations This can help you to determine if a method is over allocating objects in your code Note This view is not available if you have set the Investigate By option to Heap Count or Heap Memory in the Instances view For more information see Understanding the Instances View on page 139 When you are investigating by Recorded Count or Memory the view displays a list of methods that have allocated instances during data recording and were still live when the snapshot was taken Note during data recording In order for the Call Traces view to be available you must have collected trace information This section contains the following information e Opening the Call Traces View on page 175 e Understanding the Call Traces View on page 175 Exploring Memory Analysis Views 175 Exploring the Call Traces View e Customizing the Call Traces View on page 177 e Interacting with the Call Traces View on page 178 Opening the Call Traces View If you took a snapshot with stack traces during a session you can view recorded and garbage collection data in the Call Traces view To open the Call Traces view 1 In the Snapshots Navigator double click the snapshot you want to display The Instances view displays the object data
223. ed in the specified location For more information see Runtime Summary Report on page 204 Create a support bundle Click Create Support Bundle alto automatically collect basic support information into a ZIP file for submission to Quest Support The Support Bundle Created dialog box appears indicating the location where the ZIP file was saved You can click Copy Bundle Name to copy to the clipboard the name of the bundle that was just created Display non heap data Click the Configure Pools button to open the Select Memory Pools dialog box then select the Non Heap option Note This action may not be available for all JVMs The graphs refresh to display the non heap data Information on other data structures that are occupying memory such as the permanent generation may be included if the JVM supports it Exploring Memory Analysis Views 137 Exploring the Memory Runtime Summary View Task Change the data interval for the memory pools and garbage collections Action Select the appropriate option from the Pools and GC Interval list Response The Memory Pools and Garbage Collection graphs change to the selected interval Save the chart as a PNG or JPEG file Right click in the chart and select Save Chart As then select the graphic file type PNG or JPEG in which you want to save the chart In the Export Image dialog box specify the export paramete
224. ed methods methodZz is for your calling method and attributes it to the aggregated and is inflated with calling method In the attributed to performance data for Call Graph the methodA methods that you may encapsulated methods In the Call Graph have no control over are not displayed method calls methodB 270 JProbe User Guide Policy Description Call Sequence Advantage Example Disadvantage Grouped JProbe does not collect Data is collected Using this policy the data for encapsulated for method data calling method s data is methods In the Call for methodx preserved and you can Graph calls to one or methody and see that the method more encapsulated methodZ is made at least one call methods are represented aggregated into an out to an encapsulated by a single link toa asterisk node In method The names of node containing only an the Call Graph the encapsulated asterisk methodA calls an methods are lost but asterisk node you are not interested in which in turn calls these methods anyway method Full For each encapsulated The data for This policy provides the method called by a non encapsulated method JProbe aggregates the data for the encapsulated method plus any other encapsulated methods that it calls In the Call Graph these nodes contain a lock icon methodx methody and methodz is aggregated and attributed to methodx In the Call Graph methoda call
225. ed x4 Column Data Description Parent Method The name of the parent method for the CCT node Calls The number of times the parent called the current method for a CCT node CPU Method The CPU resources used by the current method when called by its parent Elap Method Displays the method time for the current method when called by its parent CPU Cml Displays CPU cumulative time when called by parent Elap Cml Displays cumulative time when called by parent JProbe User Guide Column Data Description Child CPU The contribution by the child method to the child CPU time of the current method Child Elap The contribution by the child method to the child elapsed time of the current method Calls to Child Displays the number of calls to the current method by the parent method Child Method The name of the child method For more information about recursive calls see Recursive Methods on page 232 Navigating the Method Detail View Use the Method Detail view to navigate through the call stack and view the current method s performance data The Method Detail view behaves similarly to the history navigation of a web browser You can use the forward button to move to the next method and use the back button to view previous method detail method Double click the parent in the Edges From Parents or child in Edges To Children to make that method the current method The following t
226. eeosst eens susseersergeesesanrsstierueseteaeraseuatver 384 Opening the Source Code rst a iaa aA kan ahai amanda tinea cide 384 Understanding the Data in a Source VIEW c csesssessesssssessessessessessessessessesessssenceusseseuesansaeeaesaneaneatsateateateeteaseess 385 Interacting with the Source VieW s sc0c secs 2cnevaienraecwnavari a E a EEEE REN 388 Setting Preferences for Coverage VIEWS sscssessesessessesnsestesseesneessesseesesseeseestesseesecsiesnsesusaneesteentesnsenteeneeaeeneennees 390 Displaying Percentages in Bar Charts csssessecessecssecsssecsseseesecseesseesecseesneeasecntesseeaseenseaseesseeseeseestesneensenses 391 Filtering Out Catch Block Datta D 391 Customizing the Colors Used to Display Results cssscsssesstesseseeseseesesseseesesseeseesteestesnseseeneeeeeneentennees 392 Learning More About Coverage Results essessessecsssecssesseestecstessesseeseesseesecseesueeseeeneeaeeaeesueeateesteeseeseestesneenseeaes 393 Nonintuitive Results for Some Source Code essesessesecsseseestessteseesesseeseesueestesnseseentesnseaseeneeseesneentesneenenaes 393 Nonintuitive Results for Conditions ccssescsssssssscssessscessssscsessssscessssscessssscsssscasscssessesesessscseessarsesenseesscnenstevsenseess 396 12 JProbe User Guide Reporting Coverage Analysis Results s ssscssssssssssssssssesssesesssesesssesseseesesnsesesseeeeeaeenseseeaeeatentsatenteaneetnanees 399 Coverage Session ReDOM G
227. eing out of memory you may need to use this option Note Unless the Data Recording is off a new recording session always begins after you click OK even if you did not change the level Performing a Basic Memory Analysis Before you can run a Memory analysis you need a settings file and a script file that starts your application For a description of how to produce these files see Configuring JProbe to Analyze Your Application The next steps in a JProbe Memory analysis are to start your application and then connect to it from the JProbe Console You can then analyze the data that JProbe collects and displays in the Runtime Summary view Running a Memory Analysis 113 Finding Memory Leaks To run a JProbe Memory analysis 1 From the command line run the batch file that was created when you created your settings file Note This procedure offers only an example of how you can run a Memory analysis using JProbe The actual steps may vary depending on your specific application 2 In JProbe click Attach to Session i on the toolbar The Attach to Running Session dialog box appears Attach to Running Session ines JMX Select the engine or enter the Host Name and Port Number below the table Name Product User com sun enterprise glassfish bootstrap ASMain Memory azafar Memory Memory lupas Host Name IP Address localhost Port 52991 OK Cancel 3 Select the ap
228. elect code from an element such as a WAR EAR or JAR file click the cell browse button and navigate to the element When you select the element you can expand its contents Locate and double click the desired code e Enter the package class or method no signature in the cell For more information see Syntax for Filters on page 406 The first cell in the row indicates whether or not the filter is enabled 2 Set the action to Include or Exclude the code from the report Tip You cannot interact with the Action cell until you specify a method 3 Add additional filters If two filters refer to the same code the filter lower in the table is used 4 You can change your list of filters e To move a filter higher or lower in the table drag the filter to its new location e To disable a filter deselect the checkbox e To delete a filter select the row right click and select Delete 406 JProbe User Guide Syntax for Filters Filters are case sensitive and blank spaces are not allowed You can use an asterisk as a wild card it matches any character including the package separator If you specify the code in a shorthand or partial form JProbe automatically changes it to canonical form The change may not be exactly what you desire so be sure to verify it JProbe does not check to see whether the filter actually matches anything in your code For example if you make a typing error in your package name nothing in your code
229. elp 2 Define the following run configuration settings e Select the JProbe script to start your application server The startup script field is pre populated with the latest startup script saved for the selected configuration Note Ifthe startup script file needs any environment settings you should set these environment settings before executing the jpconfig or the jpconsole application For example if the script file needs a JAVA_HOME value you should first set this value in your environment then run jpconfig or jpconsole and finally proceed to run the startup script file e Optional Type in any script parameters that should be used when launching the script e Select your working directory where the startup script resides e Optional Select the Close syscon window on exit check box to automatically close the jpsyscon window after the termination of the launched script if the program exits normally If this check box is cleared or if the program exits with a non zero status you must press ENTER to close the jpsyscon window Note This option is global and is not saved against each configuration value 3 In this screen you have the following options e Run your application with JProbe Click Run The script is launched automatically and you can start to analyze your application right away For more information see Running Your Script on page 55 e Run your application with JProbe and connect the JProbe Console
230. em time cannot be reported as fractions of this unit Instead the time for these small methods lines of code will be reported as zero Measuring Performance Improvements After you solve a performance bottleneck you can measure how much of a performance improvement you gained For example you probably want to know if the algorithm itself is faster You may also want to understand what impact the fixed algorithm has on the entire program Note To ensure you are comparing similar data use the same settings file for both sessions Remember to run the improved code a few times to allow optimizations to occur The tool for measuring performance improvements is the Snapshot Difference view For details see Exploring the Performance Difference View on page 338 There are two approaches to measuring performance improvements e Measuring Improvements in a Modified Algorithm e Assessing Overall Program Performance Measuring Improvements in a Modified Algorithm You can compare the performance snapshot for the modified code with the performance snapshot taken of the original code The result will be a method by method summary of the increase or decrease in method times objects Decreases represent performance improvements and are shown as negative numbers Note The JVM you use can influence the performance of your program over time as optimizations are made If you ran the original session a few times before taking the baseline snaps
231. ening or closing brace Other factors can affect the total For more information see Nonintuitive Results for Some Source Code on page 393 The condition results for a line are not what expect The compiler can introduce conditions into the bytecode that do not explicitly exist in the source code These hidden conditions affect the total number of possible outcomes For more information see Nonintuitive Results for Conditions on page 396 An EJB class appears twice in my results Some J2EE application servers such as BEA WebLogic 8 1 may load EJBs twice using different classloaders JProbe treats these instances as different EJB classes and tracks Coverage data separately 412 JProbe User Guide 21 Learning How to Use JProbe Effectively This chapter provides background information on how JProbe works and suggests some things that you can do to use JProbe effectively This chapter contains the following sections Hints for Success with IPOD c ccccccccssecssceccsssscsvsesssonsnececiassocsececsvsnsccesecsesssdensuunseevesssevtneees 414 Learning How JPiObe Woke casandsenkanddea sidvsinaaidw dae a 416 Perms ouno iE EEE oie a 418 A Quick Tour of the Interface oo ccc cc ccessscscscesseesscscesesesessusssessceceuesesessusneneecesevenenenens 419 414 JProbe User Guide Hints for Success with JProbe The more familiar you are with the application that you want to analyze the easier it will be for you to inter
232. er that allows you to run JProbe sessions and details about the license options JProbe Documentation Suite The JProbe documentation suite is provided in a combination of online help PDF HTML and TXT e Online Help You can open the online help by clicking the Help icon on the JProbe toolbar e PDF The complete JProbe documentation set is available in PDF format on SupportLink The PDF documentation can also be found in the Documentation folder on the JProbe DVD The default location of the documentation after an installation is lt jprobe_home gt docs Adobe Reader is required HTML Release Notes are provided in HTML and TXT format The default location of this document after an installation is lt jprobe_home gt docs Introduction to This Guide 19 JProbe Documentation Suite The Ant Tasks User Manual is also provided in HTML format The default location of this document after an installation is lt jprobe_home gt automation doc To open it navigate to index html Core Documentation Set The core documentation set consists of the following files e JProbe Installation Guide PDF e JProbe User Guide PDF and online help e JProbe Reference Guide PDF e JProbe Plugins for Eclipse Guide PDF e JProbe Tutorials PDF and online help e JProbe Release Notes HTML and TXT e Ant Tasks User Manual HTML Feedback on the Documentation We are interested in receiving feedback from you about our documentation For example d
233. er of live instances allocated directly by the location The tree graph or table may group multiple nodes by category and display just the category name to help you focus on the methods that matter to you You can drill down into these grouped items by choosing them in the table or tree and clicking Replace Category with More Detail on the toolbar Exploring Memory Analysis Views 177 Exploring the Call Traces View To investigate which objects were allocated by a selected package class or method click Instance Detail View E For more information about using the Instance Detail view see Exploring the Instance Detail View on page 166 Customizing the Call Traces View You can change the memory units displayed in the method list Tools gt Edit Options gt Data Display gt Memory You can also customize your Call Traces view by changing the font family size and style as well as the color scheme e Changing the Scale Gradation on page 177 e Changing the Color Scheme on page 177 e Customizing Fonts on page 178 Changing the Scale Gradation By default the scale gradation is geometric which means that except for the first two ranges the range of the segments doubles from minimum to maximum For example if the first segment spans 15 units the second spans 15 the third 30 the fourth 60 and so on doubling in size until the maximum value is reached Use a geometric gradation to see the
234. er that was instantiated truncated to 50 characters JProbe provides several truncation options For details see Setting Preferences for Performance Views Note This is the URL parameter value not the connection object Calls Displays how many times getConnection was called with each URL CPU Cml Displays the CPU time spent in getConnection for each URL Elap Cml Displays the elapsed time spent in getConnection for each URL By double clicking a row in the Connections section you open the Methods view with the graph associated with the row as the focus node By clicking the Where Were Connections Closed button you open the Methods view with the Connection close method as the focus node Exploring Performance Analysis Views 291 Exploring the Summary View The SQL Statements section displays the time required to execute SQL statements The following table lists the type of information displayed in the SQL Statements table Parameter Information Provided SQL Represents the SQL statement truncated to 50 characters JProbe provides several truncation options For details see Setting Preferences for Performance Views Note Each instance of a prepared or a callable statement that uses a particular SQL string is represented by its own row However batch execution which contains more than one SQL statement is represented in the SQL column only by an asterisk Select the row right click and select View SQL S
235. ercentages as bar charts displaying 391 CSV export options setting 455 general options setting 454 method signatures hiding 455 editing a configuration 60 categories 436 EJB class 411 JAR file 43 elapsed time 257 elements creating filters from 43 encapsulating data on methods 250 encapsulation filtering methods 250 full option 226 251 grouped option 227 252 none option 228 253 on methods 250 options 225 policy 269 enterprise license 472 errors resolving in configurations 61 examining garbage collection data 120 execution log viewing 444 exporting data from views 442 443 F filtering Coverage reports 405 Heap Data 105 filtering data Dominating Instances view 151 Instances view 141 filters Coverage 369 creating from elements 43 data collection performance session runtime tip 352 defining for application code 38 for data collection 422 Memory 124 Performance 267 filters data collection adding 422 adding filters using My Code Packages 423 execution order changing 424 impact on metrics 241 removing 424 finding hotspots 243 memory leaks 108 folders adding 436 deleting 437 renaming 437 G g compiler option 415 garbage collection collecting data for tuning 118 examining data 120 issues 117 monitoring with the Runtime Summary view 118 tuning 117 GC Data Memory 104 134 Performance 276 Index 477 generating reports 440 generating a support bundle 446 Geronimo 76 77 78 GlassFish 87 88
236. ere conditions are added or completely removed Source Explanation of Compiler Behavior Code assert Most compilers add one hidden condition to the line for the assert which in turn adds two more possible outcomes to the total outcomes on the line class The source construct lt classname gt class is sometimes compiled into code that calls Class forName lt classname gt To avoid repeated calls the compiler generates code to check a hidden field to see if it has already saved the result of forName If it has not it calls forName and saves the result Otherwise it just uses the saved result JProbe may include this check in its total possible condition outcomes 398 JProbe User Guide Source Code debug Explanation of Compiler Behavior Some compilers may remove debug conditions For example the following source code may not appear in the bytecode private static final boolean debug false if debug System out println Debug 19 Reporting Coverage Analysis Results You can generate reports based on some of the Coverage views This chapter presents the Coverage reports and the report options To learn the steps to create a report see Generating HTML Reports on page 440 This chapter contains the following sections Goverade Session REDON so cccsiaie asad casein ane diode cine anaes 400 Snape hot Browser REDON andinki iniaa ai aa aaia daaa aeir 401 400 JProbe User Guide
237. erience a deadlock as a hang that is the program becomes unresponsive and never ends Deadlocks in Java EE applications may result in incomplete transactions or a high number of waiters in a thread pool Note In the Diner example you can see a program that deadlocks Click Help gt Content and Performance Tutorials gt Diner This behavior results from improper use of the synchronized keyword to manage thread interaction with specific objects The synchronized keyword ensures that only one thread at a time is permitted to execute a given block of code A thread must therefore have exclusive access to the class or variable before it can proceed When it accesses the object the thread locks the object and the lock causes other threads that want to access that object to block until the first thread releases the lock Since this is the case by using the synchronized keyword you can easily be caught in a situation where two threads are waiting for each other to do something A classic example for a deadlock situation is as follows class Deadlocker int field_1 private Object lock_1 new int 1 int field_2 private Object lock_2 new int 1 public void method1 int value synchronized lock_1 synchronized lock_2 field_1 0 field_2 0 public void method2 int value synchronized lock_2 synchronized lock_1 Frelaed S00 field_2 0 262 JProbe User Guide Consider this situation
238. es Set by Keep Alive Size Name Recorded Count Recorded Memory ytes Heap Count Heap Memory Bytes Keep Alive Size Bytes Jave ng tnteger 35 5 840 35 5 840 5 840 Dominating Instances by Keep Alive Size Recorded Count Recorded Memory Bytes Heap Count Heap Memory Bytes Keep Alive Size Bytes 256 4 096 s 4 096 4 096 7S 1 200 7s 1 200 1 200 17 272 17 272 272 6 96 6 s 80 s 80 80 4 e 4 s 64 Textinfo 1 16 i 16 16 1 pit Leakample 16 1 16 16 Pinned Instances Report This report contains the content of the table in the Pinned Instances view Before generating this report sort the table by the metric that you are interested in 206 JProbe User Guide Dominating Instances by Keep Alive Size Name Recorded Count Recorded Memory Bytes Heap Count Heap Memory Bytes Keep Alive Size Bytes Java utl HashMapsEntry 82 21 168 82 21 168 17 728 Pinned Instances by Keep Alive Size Name Recorded Count Recorded Memory Bytes Heap Count Heap Memory Bytes Keep Alive Size Bytes Javaeng strng 7 1 128 724 17 304 437 490 char J 62 3 468 565 27 564 27 584 153 60 056 10 424 129 3 096 130 3 120 3 296 us 2 760 115 2 760 3 166 42 1 680 42 1 680 1 680 21 s8 z 880 1 536 32 1 280 32 1 280 1 280 736 736 1 136 sunma LauncherSApoCassLosder 1 7 856 Allocations Report This report contains the content of the table in the Allocations view Before generating this report sort the table by the me
239. es code at the end of a block to skip subsequent blocks such as subsequent catch blocks or the else portion of an if statement Exploring Coverage Analysis Views 395 Learning More About Coverage Results Nonintuitive JProbe Results Data is assigned inconsistently to the closing braces of methods Explanation of Compiler Behavior In bytecode methods must always have an explicit return statement In the Java language however the return statement is optional for void methods For most compilers if a return statement exists the bytecode is assigned to it Otherwise the bytecode is assigned to the closing brace A catch line with no source code is reported as missed rather than unexecutable Most compilers generate code to save caught exceptions This code is usually but not always attributed to the catch clause itself rather than the block associated with it whether or not that block is empty Many test cases do not exercise this generated code so JProbe reports it as missed Workaround To remove all data associated with catch blocks including additional bytecode generated by the compiler choose the Filter out catch blocks display option For more information see Filtering Out Catch Block Data on page 391 The start or end of some synchronized blocks are unexpectedly colored as catch blocks when the JProbe Filter out catch blocks display option is enabled The bytecode to a
240. esneeseenseseeseesteaseesesnseetsneeanenseateneeans 214 pemon Molt Mosi 3 svacsiaa sui ead aired sain en ans enna 215 212 JProbe User Guide Memory Troubleshooting Tips The following tips can help you with problems you may encounter See the ReleaseNotes htm file for other product related issues l am having difficulties configuring my environment Contact Quest Support for assistance When you do be prepared to describe your environment The customer service representative may ask you to run a JProbe session that includes some debugging options to describe your settings more fully The JProbe Console cannot open a large memory snapshot The first time JProbe opens a memory snapshot in the JProbe Console it indexes and sorts the data in the snapshot This can be a time and memory consuming operation especially for large snapshots If you encounter a JProbe Console error try using the jpprepare utility to preprocess the snapshot For more informations see jpprepare under Utilities in the JProbe Reference Guide The session hangs when running an application server under load If you notice excessive heap growth without garbage collections followed by an application crash increase the Xms and Xmx parameters or lighten the load so that the JVM has a chance to catch up and do a garbage collection As the JVM caches data you can increase the load Note For some JVMs you will need to specify the initial heap size and
241. esnesteasateateateaneateaneers Querying a Seton INStan CS cient de cise iatiacs a in a aceasta nate Interacting with the Instance Detail VioW s sscsecsessessessessesseeseesesesessecsestecueeseeseessenseseenseateseesteaseseeatenteaseeseese Exploring the Call Traces View Opening the Call Traces ViOW 0c2t csninceiiensiiauraien aniwniwnnanianeiitecabanaereienie anaes Understanding the Call Traces View cessestssecsessesstecssesecsuesseesseesecseesueeatecnteateeaeesneeaseentesseeseeseesseeatesneensenies Customizing the Call Traces VieW arwicdscnscuaienraenasinaawideniiai ied sien latest ENEE Interacting with the Call Traces View cssesseseesteseestesseseesesseesneesesstesnseseeneesueeaseeneesneesieseenesseeseeniesneennenses Exploring the Merged Allocation Points VieW sssssssssssssssessessessesssssssesnssssnseesstenseateaseatesssaneanenneaneanenteanenens Opening the Merged Allocation Points View sssssesscseeseseeseseeseseseesesseeseeseseeseesesnseensenteenseeeeneennees Understanding the Merged Allocation Points View Interacting with the Merged Allocation Points VieW sssssssssssssesssstssestesesnssessessnssassateneeateateneeateateaneateaneers Exploring the Heap Graph VieW sinnini nii e iea i i Opening the Heap Graphiswctsiva ihe ce a ee a N R Understanding the Heap Graph ceseessecssecssssecssesnssssecsessneesesseesueessesseeseestecnteaeeatesueeasecntesseestesseesueeatesneensenaes Cu
242. ess by specifying advanced JProbe settings Click Next and continue on step 8 e Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 e Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 Optional Specify the JProbe advanced options For more information see Specifying Advanced JProbe Options on page 46 In this screen you have the following options e Continue the configuration process by reviewing the settings Click Next and continue on step 9 e Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 e Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 Optional Review the settings in the configuration summary screen For more information see Reviewing the Configuration Summary on page 49 The configuration is now complete In this screen you have the following options e Edit any of the configuration settings if necessary Click Edit beside the setting that needs further updates The wizard opens the screen correspondi
243. essing JProbe analyzes the paths in the heap and identifies which instances are reachable only through other instances It builds a tree structure called a dominator tree which starts with the roots of the heap and includes all instances in the heap Using the Dominating Instances view you can analyze why an instance is live by navigating towards the roots of the tree The Dominating Instances view consists of three panes 150 JProbe User Guide Welcome instances lei Dominating Instances Pinned Instances by Instance Type Name Recorded Count Heap Count Recorded Memory Heap Memory Keep Alive Size Y java lang Class 2 038 849 560 1 380 392 Pinned Memory by Dominating Type Categories I lt IvM Process Roots gt i lt sticky Class Root gt Bava se mos memory leakexample LeakExample r E lt Other Root gt o 20 40 60 80 100 Pinned Instances by Dominating Type Filter Classes e 1106 7 1061 Name Recorded Count Heap Count Recorded Memory Heap Memory Keep Alive Size lt JVM Process Roots gt x 1 869 3 780 424 1 251 008 lt Sticky Class Root gt 49 a 29 032 108 712 java lang Class 16 5 552 65 552 java lang reflect Constru 1 432 2 064 java lang reflect Method 1 328 1 472 sun reflect GeneratedMet 1 344 912 sun reflert GeaneraterSeri 1 344 912 M Investigating Heap Instances Immediate
244. essssesesseseesessessesssesnssessserssssessetsseneesneateaneseeasssenseteatenseateateaeeases 73 Runtime Notes for Application Servers saccc ccnsencnea Gn ratvaier acne vidivasiniaiaraneeteainninAanaes 75 Apache Geronimo V1 0 s 0ceaulienad vari nace as dana a 76 Apache Geronimo tendaccis arate ditties eeiralea dled R tn Media 77 Apache Geronimo v2 O EAE EE AAE A T E ER T 77 Table of Contents 5 Apache Geronimo V2 sisters eit e a daar lial atl 78 Apache Tomcat V0 0 2 as satan ada a eae veered ee ae rete 79 Apache Lomcat v6 0 3 ihscs nck eats cal eee ae ode te en ee abs a tial 80 BEA Systems WebLOGICVOX B Xi icdssescsetsiaacsnessecsaseicasantansans dee sa AEE aaaea pahanda E Eaa CARPI naan 81 BEA Systems WebLogic Portal V9 X niie a caries cove awe AEEA A A 81 BEA Systems Weblogi V10 0 a sicsatecnnasraanniaraacieavinnas nn anidiaaiamentanie anaanieranaraniens 81 IBM WebSphere AS Community Edition V2 0 0 1 cssesssessessessessessessessessessessssenssscenseussaeeneeassaeeaesaesaneanesteaneaseesees 82 IBM WebSphere AS Community Edition V2 1 oo ecesesseseestessesseeseesneesnseseeseesneensesstessesseeseesessesnsesneeteeseantesneens 82 IBM WebSphere ASVO X rra iprasta ee aaaeei AAAS EACEA E NEOS AASE ATARE Aaaa AEAEE E Aa EATA OA ESEESE a 83 IBM WebSphere Portal Server v6 x IBM WebSphere AS V7 0 ssssscsessessessessessessessecsessessnsaessnsenssnssussussussussussussuseussaseuseussaeeaeeansatsansusseseeseaneaseasea
245. etal MetalButtonUI OXa6d6 13 2 false jJavax swing AbstractButton Handler OX22aa Fs 1 true Memory Snapshot Difference Report This report contains the contents of the Memory Difference view Before generating this report sort the table by the metric that you are interested in Reporting Memory Analysis Results 209 Memory Snapshot Difference Report Snapshot Difference by Heap Count Use Case Use Case Heap Heap Memory Dead Dead Memory Count Memory Bytes Count Bytes Count Bytes o 24 24 Name char java lang String sun java2d windows GDIBlitLoops java util Hashtable java awt image VolatileImage sun java2d loops DrawGlyphList Java nio charset CoderResult java util concurrent locks ReentrantLock NonfairSync java io ObjectStreamField java io Console 1 1 ce00000000 e00000000 oooo0oo0oo0oo0o0rH oooooo0oo0oo0 ce00000000 osgogoscogose 210 JProbe User Guide 8 Troubleshooting Memory Analysis Results As you work with JProbe you may need help to get the most out of your analysis sessions This chapter presents a few tips for solving some Memory analysis problems and for running sessions efficiently For Memory analysis issues that are not covered here refer to Quest s technical support This chapter contains the following sections Memory Troubleshooting TIPS isis ciciiciidb inane ndaadnaonanindnannwaddun 212 Memory Session Runtime Tips c scecscsecssesseestssteetesneetesnset
246. ethod Where more than one series leads to the allocating method each series is shown as a branch in the tree The value in the Count column for the allocating method is the total of all branches If the Merged Allocation Points view shows Empty dataset No matches either the class has no instances of traces or allocation point data was not collected at the time that the instances of the class were allocated Stack traces are only collected on instances that were allocated after you started recording data Interacting with the Merged Allocation Points View You can expand and collapse trace entries as well as review the instances allocated by the method by opening the Instance Detail view Where you have access to the source code you can also drill down to the code for either the allocating method or the selected class If the Source column is blank and or the Allocated At Source option is disabled Exploring Memory Analysis Views 183 Exploring the Merged Allocation Points View you do not have access to the source code for this method If the Class Source option is disabled you do not have access to the source code for the class The following table summarizes the tasks that you can perform Task Expand the tree Action Select a node right click and select Expand to Next Branch Point Response The node expands to the next node Examine the instances allocated by a method Click Instance Detail View be
247. ethod may have called other methods that are not shown in the graph This happens for native methods and for methods collected while the encapsulation level is set to None In this case you can view the list of methods represented by this node by selecting the node right clicking and selecting View Method Names For normal method nodes captured without encapsulation the View Method Names list represents encapsulated methods that the selected method may have called Those encapsulated methods may have also called other encapsulated methods The Method Time calculated in this case for the normal method node includes the named method and possibly time from none some or all of the listed methods and other encapsulated methods they called For grouped nodes the View Method Names list represents encapsulated methods that the parent method may have called The Method Time calculated in this case for the grouped node includes time from some or all of the listed methods and any encapsulated methods they called 302 JProbe User Guide Icon pa J Type Encapsulated Grouped Description These types of nodes indicate that a method is encapsulated The lock represents an encapsulated method The lightning bolt represents a simple method The asterisk represents a method that is encapsulated using the group type Details about calls to other methods or from other methods are not shown but their time a
248. ew 149 garbage collection 117 Heap Graph view 186 Instance Detail view 167 Instances view 139 Leak Doctor view 192 Memory Difference view 199 Merged Allocation Points view 181 Method Detail view 326 Performance Difference view 339 Performance Methods view 297 Performance Summary view 285 Pinned Instances view 155 Runtime Summary view Memory 132 Performance 274 undoing navigate actions in the Call Graph 323 unloading snapshots 449 unsaved snapshot warning disabling 454 unsupported application server creating a configuration for 65 modifying a startup script 68 requesting modules from Quest Support 67 use cases selecting 414 V viewing execution log 444 session history 445 views Allocations 160 Allocations Detail 163 Call Traces 121 174 Coverage Session 374 Coverage Source 384 Dominating Instances 148 for Coverage analysis 373 for Memory analysis 131 Index 485 for Performance analysis 273 Heap Graph 185 Instance Detail 166 Instances 138 Leak Doctor 190 Memory Difference 197 Memory Source 196 Merged Allocation Points 120 181 Method Detail 325 Methods 297 Performance Difference 338 Performance Source 335 Pinned Instances 155 printing 441 Runtime Summary Memory 132 Performance 274 Summary 283 W WAR file 43 warnings resolving in configurations 61 Web browser changing 454 WebLogic 81 WebLogic Portal 81 WebSphere AS 83 85 WebSphere AS Community Ed 82 WebSphere Portal Server
249. f trace information was collected for the selected classes To open the Allocations view e In the Instances view click Show Allocations View F The Allocations view appears Understanding the Allocations View When analyzing your application s memory usage it is beneficial to view data from several different perspectives If you noticed steep increases in the heap usage at certain points a method might be allocating a large number of objects The Allocations view offers you a different way of viewing object allocations This can help you to determine if a method is over allocating objects in your code By default the view displays a list of methods that have allocated live instances during the use case Once you have established which method is over allocating objects you can further your investigation by viewing related objects using the Location Detail view The Allocations view provides you with multiple ways of viewing the same method data The Group Data By list provide you with a choice of how you want to view data It offers the following options by Package Class and Method Exploring Memory Analysis Views 161 Exploring the Allocations View Interacting with the Allocations View The following table summarizes the tasks that you can perform Task Action Response Investigate objects allocated by the selected package class or method Select a package class or method and click Show Allocatio
250. feature allows you to distinguish between the time spent in their code and time spent working with the database component Common database drivers and those that ship with supported application servers are recognized by default When launching a Performance analysis with JOBC component instrumentation enabled JProbe instruments the JDBC driver classes and treats them as a component boundary The methods invoked from the JDBC classes are presented in the call tree as JDBC nodes with additional data Filters for Performance You can select Data Collection Filters for your Performance analysis in the Filters tab of the configuration wizard Filters help you to focus on those aspects of the program that are important to you Your Data Collection Filters table may already contain a set of default filters The default filters are based on the description of your code that you provided see the My Code tab in the Create Edit Settings dialog box You can use these filters or edit them Filters are defined in the same way for all the analysis tools however the filter actions differ among the analysis tools This section summarizes the filter actions for a Performance analysis To learn how to create filters see Refining Data Collection Filters 268 JProbe User Guide Filter Action Method Description Default Collects method level data on the specified code Select this action for code that you are interested in analyzing Encapsu
251. file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Create a report for this view Click Export Data to HTML a The Create Heap Graph Report dialog box appears For details see Heap Graph Report on page 208 190 JProbe User Guide Exploring the Leak Doctor View After you have identified a potential loiterer the Leak Doctor displays the references to it as a tree You can explore the paths and model what would happen if you freed a reference along the reference path If freeing a reference results in the loitering object being released you know where you have to fix your code This section contains the following information e Opening the Leak Doctor on page 190 e Understanding the Leak Doctor on page 192 e Interacting with the Leak Doctor View on page 195 Opening the Leak Doctor You use the Leak Doctor view after you have identified your loitering object To open the Leak Doctor 1 In the Instance Detail or Heap Graph view click the loitering instance or Select a group of instances in the Instance Query view 2 Click Leak Doctor The Leak Doctor view opens displaying the references to the loitering object Exploring Memory Analysis Views 191 Exploring the Leak Doctor View Welcome I instances ll Dominating Instances Da instance Detail
252. formation e Opening the Source View e Interacting with the Source View Exploring Memory Analysis Views 197 Exploring the Memory Difference View Opening the Source View The Source view displays the source code for the method that allocated the instances of a class If you have the source code for a class you can open the Source view from the following views Instance Detail Call Traces and Merged Allocation Points To open the Source view and display class source code 1 Click a class or instance of a class from a view that has access to the Source view 2 Right click the selected class instance and select Allocated At Source The Source view opens with code for the selected class displayed Interacting with the Source View You can only view your source code in this view you must edit your source code outside of JProbe Exploring the Memory Difference View The Memory Difference view summarizes the differences between two related memory snapshots The differences are expressed in terms of the class level metrics available in both the Runtime Summary and the Instances view Recorded Heap and Dead counts and memory Note To get meaningful results you must compare similar memory snapshots This means that the snapshots need to be taken at the same relative point in each session This section contains the following information e Opening the Memory Difference View e Understanding the Memory Difference View e Using Fi
253. fully errors are saved to lt WEBSPHERE_HOME gt Appserver profiles lt PROFILE_NAME gt logs lt SERVER_NAME gt native_stderr out Caution Do not end this script using Ctrl C It leaves your application server in an unstable state Run the JProbe start script with clean argument if an abnormal termination occurs WebSphere Profiles If you want to run JProbe with a WebSphere profile you need to set the Server Bin Directory field in the configuration to the profile s bin directory You do not need to set the profilename lt profile gt option JBoss v4 2 x You run the JProbe generated startup script in the same way that you run your JBoss startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables such as JAVA_HOME and JAVA_OPTS and the same command line arguments Usage lt configuration_startup_script gt options where options are e c configuration lt name gt Set the server configuration name Starting Your Application with JProbe Attached 87 Runtime Notes for Application Servers e p patchdir lt dir gt Set the patch directory must be absolu
254. g JProbe your Performance analysis results will also differ across JVMs Due to JVM optimizations even using the same JVM for multiple sessions may cause different results You should run a few sessions under your selected JVM to allow the optimizations to occur When you begin to get fairly consistent values take your baseline snapshot If the values continue to fluctuate significantly elements of your development environment may be affecting the results Consider switching to CPU time to avoid including time spent in program pauses or while waiting for I O For more information see Changing How You Measure Time on page 257 240 JProbe User Guide You save baseline performance snapshots in exactly the same way as you save any snapshot you do not have to do anything special For convenience however you may want to consider saving the baseline snapshots in a separate directory and or with unique names For example you might create a Baseline directory and save the performance snapshots with names that reflect the sessions you ran For details see Saving Snapshots on page 448 Understanding Data Collection Before you begin identifying and investigating performance bottlenecks you need to understand the data that is collected during the session JProbe collects data for each method called by your program in terms of up to eleven performance metrics Four of these metrics are collected the rest are calculated The basic me
255. g Performance Analysis Views 309 Exploring the Methods View Task Action Response Move around in a Do one of the following The Call Graph shows the large graph e Scroll using the scroll bars across the bottom and to the right of the graph e Click Show Graph Panner t to open the Graph Panner window then move the rectangle to the part of the graph that you want to see section of the graph that you selected Zoom out to see more of the graph Move the Scale slider to the left The scale of the graph is reduced to display more nodes Zoom in to focus on Move the Scale slider to the The scale of the graph is a few nodes right increased and fewer nodes are shown Clean up the display Right click the graph and select The nodes are rearranged of the nodes in the graph Layout Graph to improve readability Save graph as a PNG or JPEG file Right click in the graph and select Save Graph As then select the graphic file type PNG or JPEG in which you want to save the graph In the Export Image dialog box specify the export parameters and a location to save the file The file is saved to the location you indicate 310 JProbe User Guide The following table lists the operations that you can perform from the context menu of a Call Graph node Task Action select a node in the Call Response Graph tab right click and select Prune a subtree Prune Selected Node
256. g a Performance Analysis 243 Investigating Your Results 2 Click Edit to change the settings 3 If the method is already listed in the table select an action by clicking the arrow in the Action column Otherwise add a new row for the method Arrange the rows in the order in which you want the filters applied 4 Click Save then Close Investigating Your Results This section includes the following topics e Identifying Performance Bottlenecks e Investigating Performance Bottlenecks e Narrowing Your Investigation e Measuring Performance Improvements e Identifying Deadlocks e Investigating Performance Deadlocks Identifying Performance Bottlenecks Your goal is to locate your most expensive methods in terms of one or more metrics As your program runs the Heap Usage Chart acts as a progress bar displaying memory use over time and reporting on garbage collections By default JProbe takes a performance snapshot when the recording level changes or when data collection ends You can use either the Call Graph or Call Tree or the method list under either Call view to find the expensive methods whichever is your preference or works best with your application Three approaches to identifying potential bottlenecks include e Finding Hotspots e Finding Hotspots in the Critical Path e Finding Hotspots by Navigating the Call Graph Finding Hotspots The method list below the Call Graph Tree is the simplest place to start You can fil
257. g from the Runtime Summary view Click Runtime Execution Log P 5 The Execution Log dialog box opens and you can review the information by navigating through the available tabs 280 JProbe User Guide Task View the history of events since the Console connected to the session Action Click View Session History fa Response The Session History dialog box opens In the left pane you can parse the list and review the events that happened since the Console last connected to the session In the right pane you can view additional details for select events Create a report for this view Click Generate Report od to open the Create Runtime Summary Report dialog box If you want to include in your report information about the threads currently displayed in the Threads tab select the Include Thread Information check box Select a location for saving the report and click Generate File The report is generated in the specified location Create a support bundle Click Create Support Bundle alto automatically collect basic support information into a ZIP file for submission to Quest Support The Support Bundle Created dialog box appears indicating the location where the ZIP file was saved You can click Copy Bundle Name to copy to the clipboard the name of the bundle that was just created Exploring Performance Analysis Views 281 Exploring the Runtime Summary Vie
258. g to the program After your application is running normally the Runtime Summary view shows the typical operating behavior The Runtime Summary view s Memory Pools tab displays heap sizes and garbage collection behavior The different pools are named by the JVM and your JVM vendor documentation may explain what they represent Usually they reflect the different generations of a generational garbage collection scheme The gauges in the lower section of the Memory Pool tab in the Distribution column show the current state of each pool The full length of the gauge represents the maximum size of the pool which may have been set directly or indirectly by command line options to the JVM for example Xmx512m The colored segment of the gauge represents an estimate of the amount of memory currently in use in the pool called Used The gray segment of the gauge represents memory that the JVM has committed from the operating system but is not yet occupied JProbe calls this Available The actual values for the Used Committed Available and Maximum memory are displayed in the columns to the right of the gauges Running a Memory Analysis 119 Tuning Garbage Collection and Finding Over allocations The chart in upper section of the Memory Pools tab shows a history of the Used portion of each of the pools along with the combined Available memory from all pools Many of the issues your vendor tuning documentation discusses are often seen cle
259. ge analysis uses the following default options Running a Coverage Analysis 361 Running Test Cases with JProbe e Analysis Type gt General tab Set Coverage Level to Line JProbe tracks which lines of code in your application are called during your test case Conditions are not tracked e Analysis Type gt Filters tab JProbe uses your application filters as the default data collection filters e Analysis Type gt Automation tab Set Initial Recording at JVM Start to Data Recording On Tip In future analyses you may want to adjust these options See Analysis Options for Coverage on page 368 c Click Save 4 To generate a startup script click Integrate a Type in a location and file name for the settings file and click Save b Click Continue Type in a location and file name for the startup script and click Save d Click Close Wizard Running Test Cases with JProbe To begin the analysis you will launch your application using the startup script created when you integrated your configuration For more information see Setting Up JProbe for a Basic Coverage Analysis on page 360 This procedure assumes that you are running manual test cases or test cases through another tool For Java SE applications if you have an input file that you pass as a command line argument you can edit your configuration to add the application argument For more information click Help in the Java Application tab At the end of this
260. h the class file If you have the source code for the class containing this method the Source view opens with the code displayed See the instances that dominate a set of instances Select a class right click and select the Why Are These Live option from the list The Dominating Instances view opens showing what instances are dominating pinning the selected instance set See the instances that are pinned by a dominating instance Select a class right click and select the What Do These Pin option from the list The Pinned Instances view is updated to display the instances dominated pinned by the selected instance set Narrow the set of instances in the view Select one or more classes and click Query 2 The Query dialog box opens Exploring Memory Analysis Views 173 Exploring the Instance Detail View Task Examine an instance and its referrers and references as a directed graph Action Select an instance and click Heap Graph Bas f Response The Heap Graph view opens with the instance selected In the graph only the instance its referrers and its references are displayed as nodes Investigate an instance using the Leak Doctor Select an instance and click Leak Doctor eg The Leak Doctor view opens with the instance selected Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Column
261. hat you can remove this reference from your code A different message may appear that indicates that no paths to a root were found for this instance This can be due to a number of reasons but does not guarantee that this instance is not being held You may want to investigate further on your own to see if this instance is contributing to the memory problem If you have removed multiple references from the graph it may be good to restore some of them to find the minimal number that are required to free the loiterer You can restore a reference by selecting it in the lower pane and clicking the N Restore Edge toolbar button Exploring Memory Analysis Views 195 Exploring the Leak Doctor View Interacting with the Leak Doctor View The following table summarizes the tasks that you can perform Task Find the next path to explore Action Select the loitering instance and click Find Next Interesting Path 5 Response The tree expands to the next commented reference Continue to do this until you locate the cause of the loitering instance Remove a reference to the loitering object Click Remove Edge s The referrer is moved to the lower pane Restore a reference to the loitering object Click Restore Edge ES The referrer is restored to the upper pane Show the Instance Details view for the selected edge Click Instance Detail The Instance Detail view is displayed Show
262. he Call Tree the Call Graph is not changed although the underlying data for the corresponding nodes does change The only exception is if the corresponding node in the graph represents the only call path in that case it is removed from the graph How Data is Affected by Using Performance Features When you use features in the Performance tool such as filtering and encapsulation pruning and isolating these actions directly affect the data in the CCT In order to understand these features refer to the Simple CCT diagram on page 223 as a starting point Filtering and Encapsulation Using the filtering and encapsulation feature allows you to view data that you want to see and minimize data that you do not want to view For example if you do not want to view data from a third party source you can encapsulate particular methods so that you are viewing data that you interested in seeing therefore creating a boundary between methods that you want to view and those that you do not want to view For example you 226 JProbe User Guide can add a filter for X encapsulated which results in X1 X2 X3 X4 and X5 to all be considered encapsulated Full Encapsulation Filtering affects the CCT that the JProbe Analysis Engine collects The JProbe Analysis Engine connects to the JVM that is executing your application and collects data Depending upon the JProbe settings that you select Collect Encapsulated the results will vary When t
263. he Configuration Summary etc You can change these default names when you save the snapshots For more information see Saving Snapshots on page 448 However you may want to be able to easily identify snapshots from sessions based on this configuration If so you can assign a base name that JProbe will use instead of snapshot during a session that is run on the basis of this configuration Note The naming convention for snapshots taken using triggers may be different A base name for a triggered snapshot can be set when the trigger is defined For more information see Setting Triggers on page 425 Reviewing the Configuration Summary Before saving your configuration review the summary of the settings that you defined i New Configuration Wizard Summary Save the Configuration Summary of the Configuration Configuration Name GamePack_Memory_WebLogicl10 Configuration Type BEA WebLogic v10 0 Server Startup Script C Program Files ApplicationServers BEA_Systems us Edit Working Directory C Program Files ApplicationServers BEA_Systems us Category Program Name Demo Analysis Type memory Advanced JProbe Options jp_output_file gamepack JProbe Port 52991 Snapshot Basename gamepack_memory JProbe Settings File Please review the summary of settings and save the configuration If you need to edit the settings click Edit Press Save amp Integrate to create a JPL file contains the co
264. he Dominating Instances VieW ssssssesesesseseseesesseseesesseeeesesseeenseesesensesseenseass 148 Exploring ihe Pinned Instancos View siusis 155 Exploring the Allocations VIEW s ssrssssieissserrersarsscinsinsuenssnnieisrernein sananen sannana aenar 160 Exploring the Allocations Detail View cssssessesesseseseesesseseeseeseseeseeseesseeseeseesseeseeatseeseeatans 163 Exploring the Instance Detail VIOW sinicssissrmniarisiniodininaiirnnninnaanienanaionanaa aa 166 Exploring the Call Traces ViGW sciri aas 174 Exploring the Merged Allocation Points View sesesssesssessessesteseetesteseetestesteneetesteneetesteneeseans 181 Exploring the Heap Graphi ViGW ccc ecteceseccsesteeceneceseesesneessterscesesnssersneseeneneatnrseestcenteteeuersie Explaining te Leak Doclor WieW a dscdcdtaGaiddan wield daw addd ans dawholiddaas Exploring the Memory Source View Exploring the Memory Difference View Setting Preferences for Memory Views 132 JProbe User Guide Exploring the Memory Runtime Summary View A Memory analysis helps you to identify references that are holding on to loitering objects and to identify the reason for this As you run a Memory session the Runtime Summary view displays instance counts and changes in counts in the Java heap After you capture the data in a memory snapshot you can use the heap investigation tools starting with the Instances view to understand the references among classes in your program Th
265. he JProbe Analysis Engine and other JProbe components do not honor this setting To change the Web browser that JProbe uses for previews of reports select User Specified In the Executable and Options fields specify the fully qualified path to the executable and any launch options respectively 3 Click OK Setting Preferences 455 Changing How JProbe Exports to a CSV File Changing How JProbe Exports to a CSV File You can select whether or not headers are included in the exported data To change how JProbe exports to a CSV file 1 Click Options iy on the toolbar or select Tools gt Options from the menu bar 2 Click CSV Export 3 To include column headings in the exported data select the Export Column Headings check box 4 Click OK Hiding Package Names and Method Signatures By default wherever a class name appears in a view the package name is prepended to it When a method is identified the method signature is appended This section describes how to change the default data display settings To hide package names or method signatures 1 Click Options iy on the toolbar or select Tools gt Options from the menu bar 2 Click Data Display 3 Do any of the following To hide method signatures clear the Show Method Signatures check box e To hide package names clear the Show Package Names check box 4 Click OK Specifying Source File Locations You can tell JProbe where to look for source files by listing t
266. he JProbe documentation suite and Quest Software This chapter contains the following sections PUN POE ise ecto E diREASRE A RHRAG E 16 About This GUIdG cccccesessssessssessssssssessssesssnsssssssssesesnsssessuesusueasssesesessseassnessscesesesssneaseseseaessens 16 JProbe Documentation Suite 2 006 iiss sii snara dine 18 TA Conveniigns erini a ai Raai aa taae aaia aiia a Ana Buns About Quest Software Inc 16 JProbe User Guide About JProbe JProbe is an enterprise class Java profiler that provides intelligent diagnostics on memory usage performance and test coverage It allows developers to quickly pinpoint and repair the root cause of application code performance and stability problems that obstruct component and integration integrity JProbe provides three types of analysis e Memory analysis allows a developer to identify and resolve Java memory leaks and object cycling to ensure optimal program efficiency and stability e Performance analysis allows a developer to identify and resolve Java bottlenecks and deadlocks to ensure optimal program performance and scalability e Coverage analysis allows a developer to identify un executed lines of code during unit testing to ensure test coverage and program correctness JProbe also offers an Eclipse plugin that provides intelligent code performance analysis and problem resolution from within the Eclipse Java IDE About This Guide The JProbe User Guide tells
267. he Methods view that is Call Graph Call Tree and Methods List from the toolbar buttons or the toolbar menu For details see the following topics e Exploring the Methods View Toolbar on page 306 e Exploring the Call Graph on page 308 e Exploring the Call Tree on page 312 e Exploring the Method List on page 315 e Editing the Data Set on page 318 e Navigating the Graph by Showing and Hiding Nodes on page 320 The interactions with and among the Call Graph Call Tree and Method List are designed in such a way that you can quickly locate nodes in other sections of the view For example you can e Switch between the Call Graph and the Call Tree while maintaining context e Find methods in the Methods List by selecting a node in the Call Graph which also selects the same node in the Methods List 306 JProbe User Guide e Find methods in the Call Tree by selecting a CCT node in the Method List which quickly finds the same node in the Call Tree Exploring the Methods View Toolbar The following table lists the operations that you can perform from the Methods view toolbar Task Action Response Show more graph Click Show More Nodes LA The Call Graph is nodes updated to display more nodes along the critical path until you reach the root Note To return to the default value that is top five nodes displayed use the Back history operation Group ungroup Click Do Not Gr
268. he configuration file For more information see Editing a Configuration on page 60 If no warnings or errors are displayed specify a location and file name for the startup script Caution Itis recommended to save the startup script in the default location Running the startup script from a different directory than the default one may generate errors for certain types of application servers for example JBoss 3 x Click Save The script file is saved in the specified location You are notified when the integration is complete Integration Wizard Integrate Integrating GamePack_Memory_WebLogic10 Integration status Checking JPL file existence done Saving current settings in JPL file done Validating WebLogic server script file done done The selected startup script file contains no call to java executable Creating JProbe script file in directory C Program Files ApplicationServers BEA_Systems user_projects domains weblogic1 Oigamepackibin Reading startup script startWWebLogic cmd to replace JAVA_HOME bin java with jplauncher exe done Please save JProbe startup script file to continue Note We recommend that you use the default script directory or original startup script directory to save JProbe startup script file Script file C Program Files ApplicationServers BEA_Systems user_projects domains weblogic 10 gamepack startWebL ogicWithJProbe cmd amp JPL file C JProbe workspaceitutoria
269. he directories containing your source code The directories are searched from the top of the list to the bottom To specify the locations of your source files 1 Click Options iy on the toolbar or select Tools gt Options from the menu bar 456 JProbe User Guide Click Source Files Click an empty row and specify the fully qualified path to the root directory of your source file For example for class file com acme Foo where the source is in c mysourcefiles main com acme F oo java specify c mysourcefiles main Enter additional directories one per row To manage the list of directories e To move a directory earlier or later in the search order choose the row and click Move Up or click Move Down e To insert an empty row above a row choose the row and click Insert e To delete a directory choose the directory and click Delete Click OK Configuring Memory Pools A memory or performance chart can display heap or non heap data You can configure the type of data included in the chart by configuring the memory pools in the Runtime Summary view If you want to change the type of memory pool that is displayed in a memory chart 1 Click Configure Pools _centioue Pools The Select Memory Pools dialog box appears Eden Space Survivor Space V Tenured Gen O Non Heap Select Heap or Non Heap Note These two options are mutually exclusive Select the check box beside each
270. he encapsulation type Full is selected the Performance analysis tool collects the first call to an encapsulated method but will treat that boundary method as a black box with all encapsulated methods it calls The structure of the tree is preserved and displays information about methods at the boundary but the tree is larger With the example and the filter above the CCT that JProbe collects has the following structure v x1 4 2 x2 5 2 e 4 2 xo 2 2 Full Encapsulation CCT Comparing the Full Encapsulation CCT to the Simple CCT diagram on page 223 note that x4 and X5 are gone in the Full Encapsulation CCT diagram because they were originally called by other encapsulated methods Their time now appears as part of the method time of the encapsulated method that calls them For example the node x4 that is gone was called by x1 The method time for X1 is now 4 nanoseconds which includes 1 nanosecond from the encapsulated x4 X5 is treated in a similar fashion with its value rolled into X2 however the call out to D is preserved The following diagram Full Encapsulation Call Graph displays how the Call Graph shows this CCT Strategies for Performance Analysis 227 Understanding Performance Data Full Encapsulation Call Graph Encapsulated Grouped When using the Collect Encapsulated Grouped option all encapsulated children of a non encapsulated method are combined into one node For example
271. he following e From the Licensing Setup installation screen click Next e From the JProbe Licensing dialog box click Close 7 For aremote setup go to the other computer that you specified when requesting the license file and repeat this procedure If the license file is not accessible from the other computer use FTP to put a copy of the license on the second computer 8 You can now use JProbe 468 JProbe User Guide Registering Licenses from the Command Line You use the jplicense utility to register node locked licenses You can also use this utility to list view replace and delete licenses For more information see the JProbe Reference Guide Note You can request a JProbe license file using any computer that has Internet access For more information see Requesting a License via SupportLink on page 465 To register node locked licenses from the command line 1 2 3 On the computer running your application and JProbe open a command prompt Navigate to the JPROBE_HOME bin directory To add one or more node locked licenses use the following syntax e Windows jplicense add lt license_file gt lt license_file gt e UNIX jplicense add lt license_file gt lt license_file gt where license_file is the fully qualified path to the JProbe license file s To extract a node locked license as a backup in case you need to reinstall JProbe use the following syntax e Windows jplicense ex
272. he following data about each occurrence of an instance Name Name and memory address Size Memory used by the object Allocation Time The time when the object was created Note The Allocation Time is only displayed for instances for which allocation data was recorded If allocation data was not recorded an allocation time of 00 00 000 is displayed Referrers The number of referrers to the instance References The number of references to other instances 168 JProbe User Guide Keep Alive Size The amount of memory that would be released if the object is removed from the heap Allocated At The method that allocated the instance or the package for the class containing the allocating method Referrers References The instance selected in the instance list forms the root of the Referrers tree By expanding the branches you can discover the path of references that attaches this instance to a live object and ultimately to the root set Loitering objects are caused by a live object inadvertently or unnecessarily retaining a reference to the loiterer for some period of time To allow the garbage collector to remove a loitering object and all its descendants you must remove all references to the loiterer When you locate your loiterer you can look at the References pane to see how many objects a loiterer anchors The Referrers tree provides the following information Name Type Keep Alive Size Field Name The instance tha
273. he session and generates a support bundle as a ZIP file for submission to Quest Support The Support Bundle Created dialog box appears indicating the location where the ZIP file was saved 2 Optional To copy to the clipboard the name of the bundle that was just created click Copy Bundle Name 3 Optional To open the Quest Support Web page click Quest SupportLink 4 Click OK to close the dialog box 24 Managing Snapshots Snapshots contain the data collected about your application If you have a snapshot you can view the data on any JProbe Console or from Eclipse if you have installed the JProbe Plugins for Eclipse You manage snapshots in the Snapshots Navigator Caution Snapshots are temporary you need to save them before exiting or they are deleted This chapter presents how to manage the snapshots in your project It contains the following sections Setting the Current Snapshot SAWING CNA e oe ai died he Gadi dvd ie ee ee nia Loading Saved SHARC cccscs cic Gree ade edana arinina ariaa Unloading SHADOWING orena aa a a ante Saving Snap Shots on EKning aaant iii Managing Snapshot TANSTONG isisisi irinin ansinda 451 448 JProbe User Guide Setting the Current Snapshot The current snapshot is the one supplying data to the views By default when the session ends the last snapshot taken is the current snapshot You can change the current snapshot to be one of the other snapshots listed in the Snapshots Navigato
274. he trade off between the amount of collected data and the impact on your application s performance The Set Recording Level dialog box is always initialized to the value you have last set To see the settings in effect in the JProbe Analysis Engine check the Session History for details see Viewing the Session History on page 445 Note Unless the Data Recording is off a new recording session always begins after you click OK even if you did not change the level Running an Analysis Session This procedure assumes that you have completed the following steps e created a basic configuration e set performance options Ideally you should also set a current performance baseline and future performance expectations This section includes the following topics e Exercising Your Program e Saving Baseline Snapshots e Understanding Data Collection Exercising Your Program To exercise your program 1 Start your program 2 In JProbe click Attach to Session ry on the toolbar 238 JProbe User Guide The Attach to Running Session dialog box appears s Attach to Running Session Select the engine or enter the Host Name and Port Number below the table Name Product User com sun enterprise glassfish bootstrap ASMain Memory azafar Memory Memory lupas Host Name IP Address localhost Port 52991 v OK Cancel 3 Select the appropriate engine from the list and click OK
275. hot you should do the same with your modified code To measure the performance improvement for an algorithm 1 Run the modified code using the original JPL settings file and take a performance snapshot For details see Running an Analysis Session on page 237 2 Inthe Snapshots Navigator load the baseline snapshot that contains data on the original code For details see Saving Baseline Snapshots on page 239 3 Right click the Performance snapshot containing data on the original code and select Snapshot Differencing Running a Performance Analysis 259 Investigating Your Results The Performance Difference dialog appears 4 Inthe Baseline Snapshot drop down list choose the snapshot with the modified code The Snapshot to Compare drop down list already displays the snapshot with the original code 5 Click OK The Performance Difference view opens 6 Find the modified method in the table and read the value in the Cumulative Time column 7 Ifyou need more detail select Tools gt Options choose Performance from the navigation tree and from Time Unit select a smaller unit of time such as microseconds or nanoseconds 8 Ifthe value is negative it represents a reduction in time spent in the method which translates to a performance improvement You have succeeded in improving the algorithm 9 If the value is positive the modified code runs slower than the original code You will probably want to try a differe
276. i ote etc 464 Bee en Lr ore E tartrate ng Serie ened enone Concurrent Licensing Enero LOE M oeei a aa ERa aE 464 JProbe User Guide Node Locked Licensing A node locked license allows JProbe to run on a predefined set of computers The standard license is for a single developer The license file contains the licensed host name and operating system the type of license purchase or evaluation and the JProbe name version and license number For evaluation or leased software it also contains the evaluation expiration date Requesting Node Locked Licenses You can obtain a node locked license while installing JProbe or you can skip this step and request a license when you launch JProbe If no license is installed JProbe checks its root directory for a valid license on startup and adds it automatically if found Alternatively you can go to the Quest Software SupportLink web site and request your license Requesting a License During Installation or from Within JProbe This procedure assumes that you are working from the computer where your application and JProbe are installed and that this computer has a graphical display and a connection to the Internet When you request a license file your hostname and number of CPUs is sent to Quest Software This information is required for the license file Note If the computer you want to license does not have Internet access or if you want to request a license for a different hostna
277. ialog box then make the desired changes Response JProbe collects the selected type of data or applies the filter specified in the table Take a snapshot Click Take Snapshot w A snapshot of heap is taken along with any data recorded It appears in the Snapshots Navigator Display runtime data in the Heap Data tab Click Refresh Runtime Data The table refreshes and displays the most recent runtime data Request a heap garbage collection Click Request Garbage Collection T A garbage collection operation is requested of the JVM Results can be observed in the GC Data tab View the execution log from the Runtime Summary view Click Runtime Execution Log A Bl The Execution Log dialog box opens and you can review the information by navigating through the available tabs 136 JProbe User Guide Task View the history of events since the Console connected to the session Action Click View Session History i Response The Session History dialog box opens In the left pane you can parse the list and review the events that happened since the Console last connected to the session In the right pane you can view additional details for select events Create a report for this view Click Generate Report od to open the Create Runtime Summary Report dialog box Select a location for saving the report and click Generate File The report is generat
278. ible modifying classes as they are loaded during the execution of the program Building a Calling Context Tree JProbe then builds a Calling Context Tree CCT with the Performance information captured during the execution of the application program The information in the CCT is not as complete as the Dynamic Call Tree DCT In the DCT each node represents a single call to a method from the method in the parent node which makes the tree complete but very large in size In a CCT each node has only one parent This allows the CCT to show a call history of the program in context or a call context for all metrics Strategies for Performance Analysis 223 Understanding Performance Data t v x1 3 2 x2 3 2 fe 4 2 x 2 2 a a x5 2 1 x1 2 2 o 1 4 Simple CCT How JProbe Builds a Calling Context Tree JProbe builds a CCT for each thread and creates a node in the tree to represent the thread To keep the data as a tree it also creates the root node and makes all the thread nodes children of the root node In the preceding CCT example there is a single root node that connects to the main node indicating a single thread of execution Each node has method time and method call numbers associated with it For example A 2 3 represents A method name 2 the method time and 3 the number of method calls Typically when viewing data in JProbe after clicking the Include Threads button you see multiple thread
279. ic s color Exploring Coverage Analysis Views 393 Learning More About Coverage Results Select Color Swatches HSB RGB Preview E Sample Text Sample Text E a Oo ample Text Sample Text 4 Choose a new color and click OK 5 Change other colors as desired 6 Click OK When you return to a Coverage view the new colors are used Learning More About Coverage Results When reviewing results in the Coverage views you may have some questions about how JProbe arrives at the data you see This section contains the following information e Nonintuitive Results for Some Source Code e Nonintuitive Results for Conditions For more information see Troubleshooting Coverage Analysis Results on page 409 Nonintuitive Results for Some Source Code JProbe provides test case coverage results on executable source code You may find instances however where JProbe appears to assign data to lines of source code that are clearly not executable such as a comment or single brace To understand how these results occur you need to look to your javac compiler When your javac compiler compiles your source code into class files the class files contain executable bytecode The compiler may also add debugging information that 394 JProbe User Guide maps the bytecode to lines of source code in the java file It saves this mapping in a line table in the class file The mapping can sometimes be surpr
280. icense e If you are not subscribed complete the subscription details e If someone else in your organization has subscribed with this license number you may not be able to license JProbe on your computer e If you want to run remote sessions be sure to add the host name of the second computer when prompted Tip To find the host name of a computer enter hostname in a command prompt or xterm session 5 After you complete the instructions on the Web site your JProbe license file is sent to your email address Save the license file to your computer 6 Now you can register the node locked license See Registering a License During Installation or from Within JProbe on page 466 Requesting a License via SupportLink You can request a JProbe license file using any computer that has Internet access Go to the Get JProbe License page URL below and follow the online instructions If the person using this copy of JProbe has not subscribed you need to complete the subscription details before you can receive a license file Alternatively instead of following this procedure you can contact Quest s technical support to request a license file To request a node locked license 1 Go to SupportLink at http support quest com 2 Select License Keys 3 Follow the instructions on the web site to request your license 466 JProbe User Guide e If you are not subscribed complete the subscription details e Ifsomeone else in your
281. id you notice any errors in the documentation Were any features undocumented Do you have any suggestions on how we can improve the documentation All comments are welcome Please submit your feedback to the following email address am docfeedback quest com Please do not submit Technical Support related issues to this email address 20 JProbe User Guide Text Conventions The following table summarizes how text styles are used in this guide Convention Code Description Monospace text represents code code objects and command line input This includes e Java language source code and examples of file contents e Classes objects methods properties constants and events e HTML documents tags and attributes Variables Monospace plus italic text represents variable code or command line objects that are replaced by an actual value or parameter Interface Bold text is used for interface options that you select such as menu items as well as keyboard commands Files components and documents Italic text is used to highlight the following items e Pathnames file names and programs e The names of other documents referenced in this guide About Quest Software Inc Now more than ever organizations need to work smart and improve efficiency Quest Software creates and supports smart systems management products helping our customers solve everyday IT challenges faster and easier Visit www q
282. ide Task Create a report for this view Action Click Export Data to HTML a Response The Create Memory Difference Report dialog box appears For details see Memory Snapshot Difference Report on page 208 Setting Preferences for Memory Views You set data display options in the Edit Options dialog box Option Memory Unit Description The options are e Bytes e Kilobytes e Megabytes Note Object size columns are always displayed in bytes The Memory Unit setting does not affect them For information about setting other preferences in JProbe see Setting Preferences on page 453 Reporting Memory Analysis Results You can generate reports based on most of the Memory views This chapter presents the Memory reports To learn the steps to create a report see Generating HTML Reports on page 440 This chapter contains the following sections Seting Report OPONE ciani ieira pease hastens aan 204 Runtime Summan REPO icii inasin ieoa aaia aae ii 204 er ONE eaa EN AS 204 Dominating Instances MENON oii isis divas nin cia enna dnianurincnunsnanandn 205 Finnod ins ances SUN cc faxcecgcadeccsndavevesscsvecess aaa AE 205 Pea nS REBUN iis Akins iad aun aan ea ba een ae 206 Allocations Detail MCG sii scien ciein delen seek Sete idee Gast Gada dese Gest eet 206 helanes Delal Repo enone aE EREATARA S SEEE 207 Call Taes REPO sunnan a a E E EEE EE 207 Merged Allocation
283. ides detailed information about the process of configuring JProbe to run your Java FE or Java SE application This chapter contains the following sections CONN QUITO OVERENE aniani inana adarna E UAA 24 Configuring a dava EE Alleah ceccsecesescsevesecesseececcerereserseeseeeteaehaeeetteesenentaneeeedenstener tenses 24 Contouring a Java SE AHAIA ci2 ix sopini aai 27 Detining a JP robe Contigua i i inansicien det annaratbes dared din aneetn 29 Specifying Your Java EE Application Server escceseseseseesesessesseeesseeseeaseetesneesesneenenseateneeatens 31 Specifying Your Jaya AD ONGRIIN isisisi naaa 31 Specifying a Java Virtual Maghihe sasini riisin iiaia 36 Speciyind VOUT COE iion irs n iiaia oiar aE EN a EATEN EEEE a anodes 38 Specifying adProbe Analysis Type ecnisccnisansiamanssann a A 45 Specifying Advanced JProbe Options sssssissessssssrrstenssssrrstensssnrrstennnnrrnttennnnnritrennsnnrrnrennnnsa 46 Reviewing the Configuration SUMMAT sisisi iiini naiai 49 Shang ihe COMMIT ANON e aiaa a aa aa aa a aa a 50 integrating Your AOPIOANOM cc eccc axes cede cxesecne deaeeasteteegee ntact odcesas essen orntectivandsneneseeretsess 51 Saving the Configuration and Integrating c sccsesseseeteseecesseetesneesteseeseesseetesneetesneetenseateneeatens 55 Ranning VOUT SONDI erii eree aaas eepe naaa aiena aaaea 55 Running Your Script and Connecting the Console ce sesesssesessesseesesseesteseetesneetesneetneneeatenes 58 Manag
284. igation Up to now you have used the default data collection options to investigate potential performance bottlenecks You may want to focus on a single call tree change how you measure time collect line level data for a few methods or change other data collection filters You can narrow your investigation in the following ways e Focusing on Threads and Methods in the Call Graph and Call Tree e Encapsulating Data on Methods e Collecting Data on Lines of Code e Collecting Data Using Triggers e Changing How You Measure Time Focusing on Threads and Methods in the Call Graph and Call Tree You can narrow your investigation by isolating a single thread or method subtree All nodes that are not part of the selected node s subtree are removed from the graph and tree and the method list metric data is recalculated to include only data for the isolated subtree You can also prune threads and method subtrees from the graph or tree In this way you can remove their nodes from the Call Graph and Call Tree and their contributions to cumulative time from the method list and focus on the performance of the rest of the program For more information see Editing the Data Set on page 318 Strategies for using the isolate and prune features include e Focusing on a Method 250 JProbe User Guide e Removing Threads and Method Subtrees Focusing on a Method You can select any method node and isolate its subtree The data will reflect
285. iler makes a determination about whether a line of code is executable or not JProbe cannot change this assessment You will need to become familiar with your compiler s behavior For more information see Nonintuitive Results for Some Source Code on page 393 When there are multiple statements on the same line the line is reported as hit even if some statements were not executed If a source line has any bytecode attributed to it by the line table generated by javac JProbe considers the line executable Multiple statements will normally all contribute bytecode to the line so execution of any of them causes JProbe to consider the line as hit For example if isBad return func If isBad is true the statement calling func will not be called but the line will be considered hit If isBad is false the return statement will not be executed but the line will be considered hit The best way to resolve this which is usually considered good programming practice anyway is to put at most one statement per line if isBad return func The total number of lines of code reported by JProbe does not match the number of lines in my source code When determining the total lines of code available JProbe includes only code that is potentially executable so blank lines and comments are ignored as well as most lines Troubleshooting Coverage Analysis Results 411 Coverage Troubleshooting Tips that contain only an op
286. ilosophe Calls 5 Exception Exits 0 Catches o Method Time Child Time Cumulative Time PhilosopherPanel ChopStick Object createPhilosopher lt init gt sinit gt CPU Method Time o CPU Child Time 0 CPU Cumulative Time 0 Edges From Parents Edges To Children Method Child CPU Child Elap Calls To Child Method Total oO 0 5 Total demos performance diners PhilosopherPanel cri o o 5 java lang Object lt init gt Calling Context Sites Parent Method Calls CPUMethod Elap Method CPUCmI Elap Cmi Child CPU Child Elap Calls To Child Child Method gj demos perfarmance diners P Understanding the Method Detail View The Method Detail view displays method information similarly to the Methods view but only for one method There are five different areas in the view Summary Graph Edges From Parents Edges To Children Calling Context Sites The Methods view displays a graphical representation of the execution of your program From here you can drill down on a specific method and view the results in the Method Exploring Performance Analysis Views 327 Exploring the Method Detail View Detail view For this reason it is important that you understand how your program works so that you understand the data presented in the methods and Method Detail views The Methods view displays some information that is captured in the CCT To obtain information for a particul
287. ilter s row The filter is ignored during the subsequent analysis Re enable a filter Select the check box beside the filter s row The filter is applied during the subsequent analysis Refining Data Collection and Presentation 425 Setting Triggers Task Action Response Delete a filter Right click the button at the The row is removed from the left of the filter s row and table select Delete Setting Triggers You can use triggers to perform actions automatically on entry to or exit from a method Actions include recording data taking a snapshot and requesting a garbage collection For example you can wait to start collecting data until your application calls a specific method You can further refine triggers by specifying when and how many times to execute the trigger based on either the number of method invocations or elapsed time Tip Triggers are required for most automated sessions If you are interested in running automated sessions see the JProbe Reference Guide This section contains the following information Addi Adding Triggers Setting Advanced Trigger Controls Changing Trigger Execution Order Removing Triggers Refining When a Trigger Executes Adding a Trigger in the Body of a Method ng Triggers A trigger is composed of three pieces of information method event and action The metho d is the method on which you want to set a trigger The event controls when
288. ilters for 405 Coverage Session exploring the 374 interacting with 376 toolbar 374 understanding data 375 Coverage Session report Class and Method Summaries 400 Class Summary 400 Coverage Snapshot Browser interacting with 382 opening 378 understanding data 378 Coverage Source view 388 Coverage Bar 385 interacting with 388 opening the 384 understanding data 385 CPU time 257 creating a test suite 356 critical path finding hotspots 244 476 JProbe User Guide CSV file exporting to 442 443 customizing Call Graph 323 Call Traces view 177 187 graph color scale gradation mapping 323 Method Detail view 333 D data collection basic metrics 240 catch counts 240 exception counts 240 filters 241 metrics 241 threaded programs 241 deadlocks definition 261 detecting 263 identifying 260 investigating 264 Threads tab 264 deleting a configuration 61 categories 437 folders 437 detecting deadlocks 263 determining coverage 356 Difference dialog for heap comparison 198 for Performance 338 Difference report Memory 208 Performance 347 displaying data in the Method Detail view 330 documentation core 19 feedback 19 suite 18 Dominating Instances report 205 Dominating Instances view 148 filtering data 151 interacting with 152 opening the 148 understanding the 149 dominators 169 dump importing 129 E EAR file 43 Edit Options dialog Coverage catch blocks filtering out 391 colors setting 392 p
289. in the legend As you drill down to more detailed information the chart and table refresh the view using only the instances in that category as the starting set You can click the back and forward toolbar buttons to move backward and forward through operations you have already performed or you can click the down arrow beside either button and go directly to any of the other operations that you have performed as shown below Similar to the Instances view you can sort the tables by any of the columns The text field above the table allows you to filter the classes that are listed in the table See Filtering Data in the Instances View on page 141 for details Interacting with the Pinned Instances View The following table summarizes the tasks that you can perform Task Action Response See the instances Select a class and click The Instance Detail view pinned by the 3 Viou fal opens with all the instances selected dominators Inistance Detail View i for the selected class listed OR Select a class right click and select the Show These Pinned Instances option from the list in the upper pane Note This action is available only if the selected classes have instances that are live in the heap JProbe User Guide Show immediately pinned instances Click Show Immediately The lower pane is updated to display only instances immediately pinned by the starting set Show all pinned i
290. inal startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Usage lt configuration_startup_script gt options where options are clean Restores WebSphere to a pre JProbe state after an abnormal termination e nowait e quiet e logfile lt filename gt e replacelog e trace e script lt script filename gt background Starting Your Application with JProbe Attached 85 Runtime Notes for Application Servers e timeout lt seconds gt e statusport lt portnumber gt e recovery e help Before starting the server the startup script updates the server configuration settings to use the jpjava executable instead of the java executable After starting the server the script restores the server configuration settings If the server does not start successfully errors are saved to lt WEBSPHERE_HOME gt Appserver profiles lt PROFILE_NAME gt logs lt SERVER_NAME gt native_stderr out Caution Do not end this script using Ctrl C It leaves your application server in an unstable state Run the JProbe start script with clean argument if an abnormal termination occurs IBM WebSphere AS v7 0 When you integrate your configuration JProbe creates a startup script resaves your settings file and also creates a jython script called jprobeUpdate py By default the settings file and the jython script
291. indow d Click OK The JProbe Console connects to the JProbe Analysis Engine or to the JVM that is running with the application The runtime view for the analysis type is displayed Runtime Notes for Application Servers The more you know about the changes JProbe makes to its copy of your startup script the more comfortable you will feel running JProbe on your application You should take some time to open the JProbe generated startup script and see what changes have been made In addition you need to review the runtime notes for your application server Caution Not all application servers are supported on all platforms For more information see the Supported Platforms chapter of the JProbe Installation Guide The following sections contain the runtime notes for the application servers Apache Geronimo v1 0 on page 76 Apache Geronimo v1 1 on page 77 Apache Geronimo v2 0 on page 77 Apache Geronimo v2 1 on page 78 Apache Tomcat v5 5 on page 79 Apache Tomcat v6 0 on page 80 BEA Systems WebLogic v9 x 8 x on page 81 BEA Systems WebLogic Portal v9 x on page 81 BEA Systems WebLogic v10 0 on page 81 IBM WebSphere AS Community Edition v2 0 0 1 on page 82 IBM WebSphere AS Community Edition v2 1 on page 82 IBM WebSphere AS v6 x on page 83 IBM WebSphere Portal Server v6 x on page 84 IBM WebSphere AS v7 0 on page 85 76 JProbe User G
292. inear Mapping e To highlight relative relationships among the nodes right click the color scale and select Geometric Mapping default Changing the Color Scheme You can select from a few predefined color schemes In all schemes the bright color represents the most expensive methods while the dull color is for the least expensive 324 JProbe User Guide method Select a color scheme that suits your preference If you want to print a graph in black and white the best scheme to use is Blue to Light Gray To change the color scheme Right click the color scale and select a color scheme Customizing Fonts Use the Font Chooser dialog box to change the font family size and style of text in the graph To change fonts 1 On the toolbar click to open the toolbar menu and select Set Font The Font Chooser dialog box opens Font Chooser Name Style Dialog xv Regular Sample Text Select a new font from the Name list Select Regular Bold or Italic from the Style list Select a font size from the Size list Your choices are displayed in the sample text below the list boxes Click OK Exploring Performance Analysis Views 325 Exploring the Method Detail View Exploring the Method Detail View The Calling Context Tree CCT is created using information captured during the execution of your application program The CCT contains information about the relationships between data that is data in the ca
293. ing Memory Analysis Views provides detailed information about the Memory views used to present memory data Chapter 7 Reporting Memory Analysis Results presents the Memory reports Chapter 8 Troubleshooting Memory Analysis Results presents a few tips for solving some Memory analysis problems and for running sessions efficiently Chapter 9 Strategies for Performance Analysis describes strategies for using the JProbe Performance analysis tool to uncover performance issues in your application data Chapter 10 Running a Performance Analysis describes how to run an analysis session and investigate your application s performance Chapter 11 Learning More About Performance Options presents how to fine tune the Performance analysis tool and how to analyze a heap dump Chapter 12 Exploring Performance Analysis Views provides detailed information about the Performance views used to present performance data Chapter 13 Reporting Performance Analysis Results presents the Performance reports Chapter 14 Troubleshooting Performance Analysis Results presents a few tips for solving some Performance analysis problems and for running sessions efficiently Chapter 15 Strategies for Coverage Analysis provides recommendations for how to use the JProbe Coverage analysis tool in your development environment Chapter 16 Running a Coverage Analysis presents a basic analysis using default Coverage settings Chapter 17 Learn
294. ing More About Coverage Options presents how to fine tune the Coverage analysis tool Chapter 18 Exploring Coverage Analysis Views provides detailed information about the Coverage views used to present coverage data 18 JProbe User Guide Chapter 19 Reporting Coverage Analysis Results presents the Coverage reports and the report options Chapter 20 Troubleshooting Coverage Analysis Results presents a few tips for solving some Coverage analysis problems and for running sessions efficiently Chapter 21 Learning How to Use JProbe Effectively provides background information on how JProbe works and suggests some things that you can do to use JProbe effectively Chapter 22 Refining Data Collection and Presentation provides detailed information about JProbe filters triggers and categories Chapter 23 Communicating Results presents how to create reports print to PDF save data to a CSV file and save charts as images using JProbe It also presents how you can view the analysis logs and the session history and how to send a report bundle to Quest Support Chapter 24 Managing Snapshots presents how to manage the snapshots in your project Chapter 25 Setting Preferences presents how to set preferences that affect how JProbe operates and displays data Chapter 26 Running Remote Sessions describes how to run remote JProbe sessions Chapter 27 JProbe License Options provides information about the license serv
295. ing information Name Type Field Name The instance that is referring to the instance that you are investigating The type of the reference The field in the referring instance that refers to the referenced instance 170 JProbe User Guide Size Memory used by the object Allocation The time when the object was created Time Keep Alive The amount of memory that would be released if the object is Size removed from the heap Traces The Allocation Stack Trace tab contains information about the instance selected in the instance list If class file contained line and source information the class name and line number for the allocating code appear in the Source column Note In order for Traces to be available you must have collected trace information during data recording The Trace table contains the following data Method Class and method names for the methods that allocated instances of the selected class Source The name of the class file and the line number of the method Querying a Set of Instances If there is a large number of instances in the view you can use a query to narrow the set of instances that you are investigating To narrow a Set of instances using a query 1 Select an instance in the Instance Detail view 2 Click Query Bi The Instance Query dialog box opens Exploring Memory Analysis Views 171 Exploring the Instance Detail View Referrer Count vee v i 3 Build a quer
296. ion History dialog presents events that have occurred since the Console has connected to the session for example the start recording stop recording snapshot taken events A timestamp and a short description are provided in the left pane for each event Additional details are presented in the right pane for the selected events For example the start recording event shows the details of the recording level and a list of filters which may appear abbreviated depending on their length when applicable Also for example the snapshot taken event shows the source of the request To view the session history 1 Inthe Runtime Session view click View Session History on the toolbar 446 JProbe User Guide Session History Session events since you last connected to the session Time Event Description __ Details for the Selected Event 04 03 20 859 Session connect Snapshot taken Snapshot taken Record Allocations Stack Traces and Garbage Data For All Instances Recording level change allocations t Snapshot taken Snapshot taken 2 Click OK to close the dialog box Generating a Support Bundle When running a session you can create a support bundle from the Runtime Session views To generate a support bundle 1 Inthe Runtime Session view click Create Support Bundle on the toolbar JProbe automatically collects basic information about t
297. ion you indicate The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location 148 JProbe User Guide Task Action Response Sort the table by a Click the column heading for The table is sorted To metric the metric reverse the sort order click the heading again Export the table to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Create a report for this view Click Export Data to HTML Er The Create Instance
298. ions Stack Traces and Garbage Data For All Instances e Record Data as Specified in the Table Below You can also turn data recording off These options are explained under Setting the Recording Level on page 110 In Memory recording data means recording information about objects that are allocated This can be done at varying levels of detail from the type of data collected allocation times allocation traces or information on garbage collected instances to the types of instances that are tracked record only information for instances of packages that you choose If you are primarily interested in viewing runtime data you can run your application and view the data in the Runtime Summary view While the application is running you can request a snapshot or a garbage collection and view that data after the analysis session is over A snapshot can contain the following types of data e The JVM s heap including all instances their references and referrers e Any data recorded for individual instances allocation times traces or instances that were garbage collected The information included depends on the recording options selected Note Small amounts of recorded data especially for large complex applications keep the JProbe snapshot size manageable and also make the process of investigation easier Runtime Summary The Runtime Summary is both your view of the Java heap and your control center for running an
299. ir JVM take a snapshot of the heap detach from the JVM and then analyze the snapshot using the views available with the JProbe Memory analysis tool While JProbe is attached to your JVM you may notice some impact on your application s performance When you detach JProbe your JVM returns to its preconnection state Connecting JProbe to a Local Application Started Without JProbe The following instructions apply if the program you want to attach to is running locally on the same computer you are running the JProbe Console Starting Your Application with JProbe Attached 97 Connecting a JProbe Console to an Application Started Without JProbe To connect a console to a local application started without JProbe 1 On the JProbe Console click Session gt Attach to Session JProbe searches for currently running engines and JVMs and displays this list in the Attach to Running Session dialog box Click the JMX Java Management Extension tab Attach to Running Session JPrabe Engines JMX You can attach to a running JVM on your machine and perform a limited memory analysis Select the JVM 776 demos memory leakexample LeakExample 3896 C QUEST_ 2 JProbe_8 3 bin ia32 lib cm jar jp_console_port 52991 jp_engine_id 1396781981 jp_connection_file C DOCUME 1 LLu 4004 com zerog lax LAX C Quest_Software JProbe_8 3 bin jpconsole lax C Documents and Settings LLupas Local Settings Temp lax2210 tmp gt OK Ca
300. irectory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Oracle OC4J 10g v10 1 3 You run the JProbe generated startup script in the same way that you run your Oracle OC4J startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables such as ORACLE_HOME JAVA_HOME and OC4J_JVM_ARGS and the same command line arguments such aS start Usage 94 JProbe User Guide lt configuration_startup_script gt Caution Do not end this script using Ctrl C It leaves your application server in an unstable state SpringSource dm Server v1 0 2 The SpringSource dm Server requires the filter string com jprobe to be added to the org osgi framework bootdelegation property in SERVER_HOMEWib java5 server profile and java6 server profile files You must manually add these filters into the files before integrating the SpringSource dm Server with JProbe For example you can
301. is Engine do one of the following e To register the license while installing JProbe select the Node Locked Per Seat Licenses tab in the Licensing Setup installation screen Note This option is available only if you choose to run a Custom installation e To register the license after installation launch JProbe and select the Node Locked Per Seat Licenses tab in the JProbe Licensing dialog box JProbe License Options 467 Node Locked Licensing 2 Click Add License navigate to and select the saved license file and click Open The license is installed to JPROBE_HOME config jprobe licenses Some of the details of the license are displayed in the table Node Locked Per Seat Licenses Concurrent Licenses licen Product Keyt Expir Add License 083 00 jprabe evaluation 2009 1 Show Valid Host Names Buy Now Obtain License Product Site 3 To view all the license details select the license and click View The License Properties dialog opens with the details displayed i License Properties 2009 11 21 unlimited license_version le 3 os windows product liprobe suite serial_number 0983 00830 sitraka license version 42 type evaluation 4 Click OK to close the License Properties dialog box 5 To save a backup copy of your license file in case you ever need to reinstall JProbe click Save 6 To end the license registration process do one of t
302. is section contains the following information e Opening the Allocations Detail View on page 163 e Understanding the Allocations Detail View on page 163 e Interacting with the Allocations Detail View on page 164 Opening the Allocations Detail View The Allocations Detail view can be reached from the Allocations view To open the Allocations Detail view 1 Select a package class or method in the Allocations view 2 Click Show Allocations Detail ja The Allocations Detail view appears Understanding the Allocations Detail View The Allocations Detail view displays a list of objects that were allocated by the selected item package class or method in the Allocations view and provides in depth information about that particular allocation You can view data that is allocated directly and data that is allocated indirectly By default this view displays cumulative data Objects allocated directly and indirectly are objects that were allocated directly Click Show All Allocated Instances H to return to the list of objects allocated directly and indirectly 164 JProbe User Guide If you suspect that one of these objects may be causing a problem in your code you can use the Instance Detail view or Merged Allocation Point view to continue your investigation of the object Interacting with the Allocations Detail View The following table summarizes the tasks that you can perform Task Action Response
303. is section contains the following information e Understanding the Memory Runtime Summary View e Interacting with the Memory Runtime Summary View Understanding the Memory Runtime Summary View The Memory Runtime Summary view is both a view of the Java heap and a control center for running your analysis session It helps you to identify loitering objects excessive garbage collections and over allocation of objects and also acts as a progress bar A green vertical line on the Runtime Summary chart indicates that a snapshot was taken either due to a change in recording level or because a snapshot was requested After you have attached to a running session of your application the Runtime Summary view appears It has three tabs e The Memory Pools Tab displays memory usage in graphic and table formats e The GC Data Tab displays garbage collection data in graphic and table formats e The Heap Data Tab displays information about the instances of classes in graphic and table format Toolbar buttons provide access to the snapshot and garbage monitoring controls The system messages area bottom pane displays any errors and warnings that occur including those that have occured before the this Console connected to the JProbe Analysis Engine Exploring Memory Analysis Views 133 Exploring the Memory Runtime Summary View QWelcome A Runtime Summary Configure Pools Pools and GC Interval One Minute v Classes Loaded 1086 Classes
304. is you want to run Memory Performance or Coverage Depending on your application server you may be required to provide additional details For details see the help provided in the user interface below each field At the end of this process you will have a settings file and a startup script For some application servers an additional file may be created to modify and later restore one of Configuring JProbe to Analyze Your Application 25 Configuring a Java EE Application the application server s configuration files If so the affected file is identified in the online help and also under Running a Java EE Application with JProbe Note If you are running JProbe in a heterogeneous networked environment for example UNIX on one computer and Windows on another configure JProbe on the computer that runs your application server This ensures that the correct syntax for the operating system is used To set up JProbe to analyze a Java EE application 1 Launch the JProbe Console on the computer where you run your application server Windows Click Start gt Programs gt JProbe gt JProbe Console UNIX Navigate to JPROBE_HOME bin and execute jpconsole Click Configurations in the Navigation area Right click the vendor of the application server hosting your application and click New Configuration or Select the vendor of the application server hosting your application and click Configurations gt New Configu
305. ising and it can vary with different compilers compiler versions javac options eg g or even the formatting of the source code For example javac may attribute bytecode to the closing brace of a void method even if the method has no return statement or may attribute it to the last line of code JProbe uses the mappings in the line table to match the coverage results for your bytecode to the lines in your source code JProbe cannot modify these mappings The following table summarizes some of the nonintuitive results that you may see in JProbe and provides an explanation of the compiler behavior that causes these results Nonintuitive JProbe Results A class declaration or its opening brace is reported as missed rather than unexecutable Explanation of Compiler Behavior This happens when the compiler generates a default constructor for a class with no explicit constructor but the constructor is not used Most compilers generate a line table for this method that attributes the code to the class declaration itself either the line with the class keyword or the opening brace Workaround To ignore the unused method in subsequent JProbe sessions add a data collection filter com example MyClass lt init gt exclude Data is assigned inconsistently to the closing braces of blocks Most compilers inconsistently attribute code to the closing brace of a block In addition sometimes the compiler generat
306. it check box to automatically close the jpsyscon window after the termination of the launched script if the program exits normally If this check box is cleared or if the program exits with a non zero status you must press ENTER to close the jpsyscon window Note This option is global and is not saved against each configuration value 4 Click Run The selected application server launches You can now connect JProbe to the running application and start analyzing your application Configuring JProbe to Analyze Your Application 57 Running Your Script Running Your Script from JProbe Configuration Tool To manually run your JProbe startup script from the JProbe Configuration tool 1 Inthe JProbe Configuration tool select a configuration from the Configurations Navigator The lt Configuration_File_Name jpl gt tab appears in the View area 2 Click Run The Run lt Configuration_Name gt dialog box appears E Run Demo Diner Performance Analysis Path of script to start a java application or server with JProbe C Quest_Software JPrabe_6 3 demos settings Diner_Performance_Analysis bat Script Parameters Working Directory ci iQuest_Software 3Probe_8 3 V Close syscon window on exit V Close Create Edit Settings tool while executing JProbe startup script Required Field Cancel Help 3 Define the following run configuration settings e Select the JProbe scrip
307. it gt methods that are executed when the class is initialized by the JVM y Simple Represents small Java methods typically accessors get set that are generally too simple for timed analysis JProbe automatically recognizes these methods Methods which include operations such as looping object creation exception handling or calls to other methods including constructors are not considered simple The time for simple methods displays as 0 Setting filters helps to narrow your investigation When you set filters and run a session you will see nodes in the call graph that represent the types of filters that you set 300 JProbe User Guide Icon 2r sim Sim runt 60 710 6 Type Line Description This type of node includes a gray box with lines indicates line level profiling information It displays line by line how the method time is spent When you place your cursor over the icon the tooltip reveals the information for the methods Method type cumulative time and number of calls Typically using this type of filter is expensive er sin Sim PS run 60 710 Method This type of node indicates the method which is being filtered and any parameters associated with this method Exploring Performance Analysis Views 301 Exploring the Methods View Icon Type Description None This type of node indicates that the Strings main 5 059 m
308. k data from calculations 1 Click the iy Options toolbar button or choose Tools gt Options 2 Choose Data Display gt Coverage 3 Choose Filter out Catch Blocks 4 Click OK When you return to a Coverage view the calculations reflect the filter setting 392 JProbe User Guide Exceptions Although it is not generally considered good coding practice it is possible that a line can have both normal statements and catch block statements Catch block filters do not apply in the followings situations Try and catch appear in the same line of code try Thread sleep 1000 catch Exception e If Thread sleep 1000 is executed coverage is 100 if not coverage is 0 Executed code and catch appear in the same line of code try System out print obj toString System out println catch Exception e System exit 1 If System out print is executed the coverage is 100 regardless of whether or not it throws an exception Customizing the Colors Used to Display Results You can change the colors used to represent hit lines missed lines missed conditions catch blocks that have been filtered out and unexecuted code To set colors 1 2 3 Click the iy Options toolbar button or choose Tools gt Options Choose Data Display gt Coverage Metric Colors Runtime Hit HE Missed Lines Hi Missed Conditions Lt No Data for Line E Fitered Catch Block Color JE Click Change beside a metr
309. karound e Initialize variable declarations int i 0 e Move the declaration below the first line of executable code e Try another compiler The last line of a catch block before a finally block is reported as hit even though the rest of the catch block may be missed Some older compilers may misattribute code from a finally block to the last line of code in a preceding catch block Nonintuitive Results for Conditions Compilation of Java source to bytecode may not always be as one might expect Compilers may rearrange conditions omit ones that are unnecessary or add ones not obvious in the source subject to some constraints by the Java language They often Exploring Coverage Analysis Views 397 Learning More About Coverage Results invert conditions so that a condition you expected to be reported as having been found true but not false may be reported by JProbe as false but not true For example the following source code if s null a else b may result in bytecode similar to the following aload s ifnonnull skip invokevirtual a goto end skip invokevirtual b end So while the source is true if s is null the i fnonnu11 opcode is true if s is not null JProbe cannot change how the compiler treats conditions For more information see A Closer Look at the Data in the Condition Detail Table on page 386 The following table lists some other common nonintuitive results wh
310. ked with an asterisk are shown by default The other columns can be added to the view by right clicking in the table to open the Show Hide Columns dialog box where you can choose other columns to be displayed JProbe User Guide Column Name Missed Classes Description Missed Classes Total Classes Missed Classes If the snapshot contains baseline snapshot data this column states the number of classes that were missed Otherwise this column will be zero Total Classes If the snapshot contains baseline snapshot data you see the total number of classes in your application taking filters into account Otherwise you see the total number of classes with methods that were called by your test case taking filters into account Missed Methods Missed Methods Total Methods Missed Methods Number of methods missed by the test case Total Methods Total number of methods in the class package application taking filters into account Missed Lines Missed Lines Total Lines Missed Lines Number of lines of code missed by the test case Total Lines Total lines of code in the class package application taking filters into account With Line Data Percentage of lines of code that have line data Missed Conditions Missed Conditions Total Conditions Missed Conditions Number of condition outcomes missed by the test case Total Conditions Tot
311. king correctly JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Usage lt configuration_startup_script gt run start geronimo_args where e run Start WebSphere CE Server with JProbe in the current window e start Start WebSphere CE Server with JProbe in a separate window IBM WebSphere AS Community Edition v2 1 You run the JProbe generated startup script in the same way that you run your WebSphere startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Usage lt configuration_startup_script gt run start geronimo_args Starting Your Application with JProbe Attached 83 Runtime Notes for Application Servers where e run Start WebSphere CE Server with JProbe in the current window e start Start WebSphere CE Server with JProbe in a separate window IBM WebSphere AS v6 x When you integrate your configuration JProbe creates a startup script resaves your settings file and also creates a jython script called jprobeUpdate py By default the
312. ks r Workstation r Application JProbe rire Engine R Con sotion Manager JVM Console a JProbe E JProbe S Snapshot JProbe running in a networked environment remote session JProbe Analysis Engine The JProbe Analysis Engine collects data on your application as it runs within its JVM The selected JVM must support the current Sun JVM profiling interface called JVMTI The profiling interface provides the hooks that the JProbe Analysis Engine needs to analyze your application The profiling interface may add some overhead so do not expect your application to run as quickly as it would without the use of the profiling interface Programs that execute very quickly are the most sensitive to this overhead JProbe Connection Manager The JProbe Connection Manager runs in its own JVM and handles communications between the JProbe Analysis Engine and a JProbe Console Only one JProbe Console can be connected to the JProbe Connection Manager at a time The JProbe Connection Manager also transfers snapshots from the Analysis Engine to the Console JProbe Console The JProbe Console provides you with a graphical interface where you can create a configuration monitor your analysis session and review snapshots of the analysis data The JProbe Console runs in its own JVM separate from the one running your 418 JProbe User Guide application Why a separate virtual
313. lated Encapsulates data about the specified code Select this action when you do not care about the performance of the specified code or you do not have control of the source code Note How JProbe collects and presents data for encapsulated methods is controlled by an encapsulation policy For more information see Encapsulation Policy on page 269 Line Collects line level data for each line of code in the specified package class or method Select this action for a method that you know from a previous method level analysis is contributing to your bottleneck It will help you to pinpoint problem lines of code Native Encapsulates data when your application calls the specified native methods Select this action for all native methods If you do not select this action some of the time spent in a native method may be attributed to its calling method You do not need to use Native filters if your JVM is Java 6 or newer Native filters are only required for older JVMs where native methods cannot be instrumented directly Note Avoid assigning this action to non native methods If you do the resulting data may be incorrect Automation Settings for Performance You can select initial recording and encapsulation options for your Performance analysis in the Automation tab of the configuration wizard Initial Recording Options The following Initial Recording at JVM Start options are available for your Performance
314. lect a node in the graph right click and select the Show Immediate Dominator option from the list The graph expand and displays the immediate dominator of the selected method Change the status of the selected node Select the node right click and select one of the Mark As or Mark Reachable As options The color of the node or the node and all nodes reachable from it changes Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden Exploring Memory Analysis Views 189 Exploring the Heap Graph View Task Resize column widths Action To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit Response The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted Export the table to CSV Right click in the table and select Export to CSV Type a file name and click Save The
315. lick and select Show Allocated At Source Note _ If this menu item is disabled JProbe cannot associate a source file with the class file The Source view opens with the source code displayed You may be prompted to locate the source code file Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden 180 JProbe User Guide Task Resize column widths Action To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit Response The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location Cannot be used on tree tables Note The column appears in its new location Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted Export the table to CSV Right click in the table and choose Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a
316. ll history Using the information from the CCT Performance can display the relationships between methods and how these relationships may affect the overall running of the application For more information see Understanding Performance Data on page 222 When using Performance to profile your program data you might want to view your data at different levels of abstraction The Call Graph provides you with a view of methods in your code at a higher level of abstraction If you are interested in investigating a single method and viewing the associations with its parent and children methods you can view this data in the Method Detail view This level of abstraction can provide valuable information especially when you are investigating methods that take a substantial amount of time to run This Method Detail view features are described in the following sections e Opening the Method Detail View e Understanding the Method Detail View e Navigating the Method Detail View e Customizing the Method Detail View 326 JProbe User Guide Opening the Method Detail View The Method Detail view is only available by selecting a method in the Methods view To open the Method Detail view In the Methods view select a method right click and select Open Method Detail View i The Method Detail view opens Welcome Runtime Summary Summary Methods View BE Method Detail T h h Current Method ChopStick lt init gt Ph
317. load the settings from the selected JPL file into a new configuration file The original JPL settings file remains unchanged Later on you can save the updated settings as a new settings file optional Configuring JProbe to Analyze Your Application 63 Managing Configurations e Save updated values into a settings file name to load the settings from the selected JPL file change any required options and save the updated settings as a new Settings file The original JPL settings file remains unchanged d Click Save amp Close The new configuration is added in the Configurations Navigator 5 Select the new configuration and click Edit a Provide a name for this new configuration b Review all the settings and edit them as necessary c Click Save 6 To generate a startup script click Integrate a Type in a location and file name for the settings file and click Save b Click Continue Type in a location and file name for the startup script and click Save d Click Close Wizard Renaming a Settings File After you create a configuration you may want to rename the settings file You can do this using the Save As button When you save the settings file with a new name the configuration is updated to reference the new settings file The startup script is not updated To update the startup script you need to reintegrate the configuration Caution If you forget to integrate you may end up running a session using an old settings file
318. lone but you can see it in the Call Graph The other drawback is that the hotspot may include data that is not relevant if it was called along different paths The Call Graph can help by showing the critical path and by allowing you to isolate or prune to focus only on relevant data By default the nodes in the graph are colored to show the most expensive methods in your program in terms of cumulative time You can easily identify the critical path of your program by following the brightest nodes in the graph The critical path represents the main path of execution for the session or sessions you ran A bottleneck in your critical path has the most impact on your customer s perception of the performance of your application Running a Performance Analysis 245 Investigating Your Results Threads are excluded from the Call Graph by default because the vast majority of threads are created by the application server You have no control over these threads so you can safely hide the data associated with them What you do see are your application s methods Your data set includes only the call trees initiated by the methods that started your Java EE application In the Call Graph you can include the threads in the data set Including threads allows you to analyze their call trees collectively and individually For example if you find an expensive method invoked by two threads you can select one thread and isolate its call tree The data change
319. loring ihe Methods VIEW aids cece ddr oeadsiiaotienddddolandiadaadohandsda an 297 Explonng the Method Detail VOW escacs 325 Exploring the Performance Source VieW sssssissessssssrrseesssssrrsrenssnrrsrenssnnnrnrinrsnnnintrerasnrnree 335 Exploring the Performance Difference ViGW s csssssssssssssssssscesesnsssetesneseenesneseenesneneeneseaes 338 Setting Preferences for Performance VieWS sssescseseesessesseeseesteseestestestestestestesseetesneesesnse 342 274 JProbe User Guide Exploring the Runtime Summary View As you run a Performance session the Runtime Summary view displays changes in counts in the Java heap After you capture the data in a memory snapshot you can use the heap investigation tools starting with the Call Graph and Call Tree to understand the performance issues in your program This section contains the following information e Understanding the Runtime Summary View e Interacting with the Performance Runtime Summary View Understanding the Runtime Summary View The Performance Runtime Summary view is both a view of the Java heap and a control center for running your analysis session It helps you to identify deadlocks and also acts as a progress bar A green vertical line on the Runtime Summary chart indicates a snapshot was taken either due to a recording level change or because a snapshot was requested After you have attached to a running session of your application the Runtime Summary
320. ls 427 algorithms inefficient 220 measuring improvements 258 Allocations Detail report 206 Allocations Detail view 163 interacting with 164 opening the 163 understanding the 163 Allocations report 206 Allocations view 160 interacting with 161 opening the 160 understanding the 160 Analysis Engine 417 analysis types specifying 45 analyzing a program using Performance 218 Apache Geronimo 76 77 78 Tomcat 79 80 application filters 38 application servers 31 Apache Geronimo 76 77 78 Index 473 Apache Tomcat 79 80 BEA Systems WebLogic 81 BEA Systems WebLogic Portal 81 GlassFish 87 88 89 IBM WebSphere AS 83 85 IBM WebSphere AS Community Ed 82 IBM WebSphere Portal Server 84 JBoss 86 87 open source 87 88 89 Oracle AS 10g 90 91 Oracle OC4J 10g 93 Oracle WebLogic 92 runtime notes for 75 SpringSource 94 unsupported 64 architecture understanding the 414 arguments specifying for a Java application 33 assessing existing test suite 356 overall program performance 259 attaching to running JVM 96 local application 96 remote application 97 attaching to session 72 73 no connection 95 automation Coverage 371 Memory 126 Performance 268 474 JProbe User Guide baseline snapshot saving a Performance 239 selecting 338 basename for snapshots 48 BEA Systems WebLogic 81 WebLogic Portal 81 bottlenecks identifying 243 narrowing your investigation 249 browser changing 454 building a C
321. ls_83 GamePack_Memory_WebLogic10_Mem_Settings jp are generated successfully You can manually run this script file from command line to start the WebLogic Server with JProbe Usage startWebL ogicWithJProbe cmd Integration complete Press Next to open Run Configuration tab for executing JPrabe startup script Close wizard Cancel Help 5 In this screen you have the following options e Close the integration wizard Click Close Wizard Configuring JProbe to Analyze Your Application 53 Integrating Your Application The Integration Wizard dialog box closes e Execute the JProbe startup script Click Next The Run Configuration tab is displayed For more information see Executing the JProbe Startup Script on page 53 Executing the JProbe Startup Script When you are ready to run JProbe on your application use the JProbe generated startup script to launch your application The startup script references the settings file and JProbe applies these settings when collecting data on your application For some application servers the startup script edits and later restores one of the application server s configuration files If so you are informed during the configuration setup For more information look up your application server in Running a Java EE Application with JProbe on page 73 If you are interested in what is happening behind the scenes you can open the startup script in a file editor Th
322. lters in the Memory Difference View e Interacting with the Memory Difference View 198 JProbe User Guide Opening the Memory Difference View You can open the Memory Difference view if you have loaded at least two memory snapshots To open the Memory Difference view 1 Load the snapshots you want to compare 2 Right click the selected snapshot and select Snapshot Differencing The Memory Difference dialog box opens The Snapshot to Compare field displays the name of the snapshot you selected 3 From the Baseline Snapshot list select the snapshot you want to use as the basis for your comparison 4 Inthe Snapshot to Compare list either leave the currently selected snapshot or select a different snapshot to compare against the baseline 5 Click OK The Memory Difference view opens ES Memory Difference Baseline snapstrot_i2 Other snapshot_i3 Percentage Size Change Use Cave 2 40 Heap 0 08 Dead 89 90 Fiter Classes w 554 564 Nome Exploring Memory Analysis Views Exploring the Memory Difference View _Use Case Count resp Count Y Use Case Memory Hesp Memory Dead Court Dead memory al Y 199 T javan sring event Changs vent hsf java lang String 1ava lang Thresd Java lang Obsect sun set windows WGobalCursoeManager java nio charset CodeR mad gt java ant Windon WindonDisposerRecord java ant Lightweghtinspstcher SUN 8v 82d ppe Loophipe java util jar Jaws
323. ly use the default Configuring JProbe to Analyze Your Application 33 Specifying Your Java Application setting that is include all the classes and later filter the list of classes to reduce its complexity Later in the wizard process JProbe creates a default filter in the My Applications Filters table for the package that contains the main class For more information see Specifying Your Code on page 38 To specify the main class 1 Specify whether this main class is a class file default or in a JAR file 2 Click the browse button beside the Main Class field The Main Class dialog box opens fF Main Class O leakexample S LeakExample 1 class Options LeakExample 2 class My Recent 5 LeakExample 3 class Documents G Desktop LeakExample class C Show only classes with main method Search JAR manifest for main class My Documents gs TOR017090 File name LeakExample class Open J My Network Places Files of type Java Classes and Containers Cancel 3 Optional In the Options tab select the check box es as necessary 4 Navigate to and select the class file or JAR file that contains the main method of your program and click Open The Main Class field displays the package and class name The Working Directory field defaults to the directory containing the package Application Arguments You can add any application arguments that you
324. ly written Snapshot C Quest_Software JProbe_8 3 snapshot_raw_1 jpp has been added to the snapshot navigator as snapshot Memory Pools Tab The data displayed in the Memory Pools graph varies depending on the JVM and the options that you are using For JVMs that use generational garbage collectors heap pools typically represent the generations For example in the Sun JVMs with some options Tenured Gen is the old generation while Eden Space and Survivor Space together make up the new generation While you are running an analysis session the chart displays the size of the heap memory pools by default Heap pools contain things normally found in the Java heap like Java objects Non heap pools normally contain things the JVM uses internally such as the compiled representation of method code or class information You can select a different time interval for the summary graph from the Pools and GC Interval list The options are One Minute Five Minutes Fifteen Minutes Thirty Minutes and One Hour The Memory Pools table shows the actual or the estimated values for each of the heap pools in the graph The column headings are defined as presented in the following table 276 JProbe User Guide Used The amount of memory currently being used Committed The amount of memory that the OS has guaranteed to be available to the JVM This is always greater than or equal to Used Available The diffe
325. m before you implement any changes to it These snapshots serve as the baseline 220 JProbe User Guide against which you assess subsequent code improvements Setting and measuring against a baseline helps you to quantify improvements to yourself your team members quality assurance product management marketing and other stakeholders Be sure to run your use case several times before selecting a snapshot to use as a baseline The JVM may perform some optimizations that show improvement only after several runs of the code For more information see Saving Baseline Snapshots on page 239 Focusing on a Known Inefficient Algorithm Any time during a development cycle a senior software developer may suspect that a heavily used algorithm is a performance bottleneck even though it is running as efficiently as possible You can use JProbe Performance to analyze how the algorithm is called For the most part however you should discourage your team from routinely using JProbe Performance on their algorithms You only want to spend time optimizing the algorithms that your customers use most frequently this use assessment is most effective when done by a senior software developer looking at the program as a whole While you run your program Performance keeps track of how long each of your methods takes to execute either in elapsed wall clock time or in CPU time The Summary view then shows you the hotspots in your program Hotspots are meth
326. machine Most importantly you can be sure that the activity of the JProbe Console does not contribute to the data collected on your program Secondly on a system with multiple processors you can take advantage of the extra processing power by running the JProbe Console on one processor and your program and the Analysis Engine on another Lastly with a dual license and remote connection this architecture allows you to run your program on a server while monitoring it from a remote workstation JProbe Snapshots Snapshots contain the data collected up to the time the snapshot is taken Whether snapshots are taken manually by you or automatically during or at the end of the session they are saved as temporary files In a remote session the temporary snapshots are transferred to the local computer and removed from the remote computer To keep snapshots you need to rename them or move them to a different location Each analysis tool has its own file format for snapshots You can distinguish among them by their icons and file extensions as shown in the following table Icon Analysis Tool File Extension Fe Memory Jjph T Performance Jjpp A Coverage jpe Demos and Tutorials JProbe ships with a set of demo applications They are located in the JProbe installation directory under demos Each demo has an inherent problem that JProbe can identify Some of the sample applications have tutorials associated with them Others are ex
327. many collections were done of each type over an interval The Garbage Collection Overhead data below the graph indicates the percentage of time the JVM was exclusively performing garbage collection compared to the elapsed time the program has been running For example a program that has been running for 10 seconds and has spent 0 5 seconds doing garbage collection has a GC overhead of 5 The lower the percentage the lower the overhead If the percentage is high you may need to tune your JVM s GC options or reduce object cycling The table provides a breakdown by garbage collector that shows the number of times each collector has been invoked and the time in milliseconds that it used as well as the cumulative counts and times Heap Data Tab The Heap Data tab is similar to the Instances view see Exploring the Instances View on page 138 The main differences are the toolbar options and the Refresh Runtime Data button The Instances view also allows more in depth analysis When you switch to the Heap Data tab during a session be sure to click this button to ensure that you are seeing the latest data Exploring Memory Analysis Views 135 Exploring the Memory Runtime Summary View Interacting with the Memory Runtime Summary View The following table summarizes the tasks that you can perform Task Change the data recording level or add a filter Action Click Set Recording Level to open the Set Recording Level d
328. me you can request a license through SupportLink For more information see Requesting a License via SupportLink on page 465 If you want to run remote sessions specify the second host name when the License Details web site is displayed The second host can only run the JProbe Console for the product not the JProbe Analysis Engine After you receive the license file you register the same license file on both computers To request a node locked license 1 On the computer running your application and the JProbe Analysis Engine do one of the following e To request the license while installing JProbe select the Node Locked Per Seat Licenses tab in the Licensing Setup installation screen Note This option is available only if you choose to run a Custom installation JProbe License Options 465 Node Locked Licensing e To request the license after installation launch JProbe and select the Node Locked Per Seat Licenses tab in the JProbe Licensing dialog box 2 Click Obtain License You are advised that your hostname and number of CPUs will be sent to Quest Software If you agree to transfer this information continue with this procedure Otherwise see Requesting a License via SupportLink on page 465 3 Click OK Your default browser opens If the Quest Software license page does not display ensure that cookies are enabled in your browser and try again 4 Follow the instructions on the web site to request your l
329. method If a critical path does not stand out do one of the following e Display a higher number of nodes by clicking Show More Nodes You can add nodes along the critical path until you reach the root e If there is little difference among the node colors in the graph try switching the color scale to Linear Mapping For details see Changing the Scale Gradation on page 323 e Select Hot Methods Elapsed Time from the Color to Show list The graph is repainted to show the methods that take the longest to execute excluding time spent in child methods Select a bright node representing one of your methods to investigate 10 After you identify one or more hotspots go to Identifying Performance Bottlenecks on page 243 Running a Performance Analysis 247 Investigating Your Results Finding Hotspots by Navigating the Call Graph When a method s cumulative time is high but the method time itself is relatively low it means that the method is calling one or more expensive methods The Call Graph shows you the calling relationships in a directed graph By navigating the graph you can find the hotspots in the call tree To navigate the graph you use the navigators on the node in the graph The navigator on the left called the parent navigator is for traversing the methods that call the selected method The navigator on the right called the child navigator is for traversing the methods called by the selected method Pa
330. mns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for the metric The table is sorted To reverse the sort order click the heading again Export the table data to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers 338 JProbe User Guide Exploring the Performance Difference View The Performance Difference view lists the differences between the current snapshot and a previous baseline snapshot To prepare to use this tab you take a snapshot of your heap data before implementing changes to fix a performance bottleneck After you have solved a performance bottleneck you can rerun the application take a snapshot and compare this new snapshot to the old snapshot to see exactly where you achieved performance gains Caution To get meaningful results you have to compare similar snapshots This means you need to use the same session settings and take the snapshots at about the same point in each session This section describes the features of the Performance Difference view e Opening the Performance Difference View e Understanding the Performance Difference View e Interacting with the Performance Difference View Opening the Performan
331. more information see Defining a JProbe Configuration on page 29 Specify the JAR file used to start your application server or the main class of your application server and the paths to the classes that the application server requires and click Next Note Specifying the JAR file for an application server is preferred as it may then use the manifest mf file to specify JAR dependencies so that you do not have to specify them directly within your JProbe settings file For more information see Specifying Your Java Application on page 31 Specify a supported JVM plus any Java options you normally use and click Next For more information see Specifying a Java Virtual Machine on page 36 Specify application filters enable JProbe to create a category containing the code set to include and click Next For more information see Specifying Your Code on page 38 Select the type of analysis you want to run For more information see Specifying a JProbe Analysis Type on page 45 In this screen you have the following options e Continue the configuration process by specifying advanced JProbe settings Click Next and continue on step 9 e Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 Configuring JProbe to Analyze Your Application 67 What Can I Do if My Application Server Is Not Supported e
332. mory Analysis 109 Finding Memory Leaks e On the stack as a local variable e Used for synchronization e Referenced by JNI code e A system class e Referenced by a thread The objects connected to the root set are either live or loitering you need to determine if the objects have future value to your program The unconnected objects no longer have any references and may be ready for garbage collection The loiterers that are easiest to detect are those that are allocated to the heap but not removed after the program is finished with them these loiterers remain for the duration of the program s execution More difficult to find are loiterers that are eventually freed or reused but in the meantime consume memory that your program could be using for other tasks Four common causes of loitering objects are e Obsolete Collection References e Obsolete Listeners e Lingering Transitional References e Stalled Stack References Obsolete Collection References An obsolete collection reference occurs when an object is added to a collection but not removed when no longer required When methods add objects without removing them the collection can grow without bounds over the runtime of the program You can avoid creating loitering objects by pairing add remove calls Obsolete Listeners Obsolete listeners are a type of obsolete collection reference However they warrant their own category because they are a common cause of loitering object
333. n JProbe tracks which lines of code are executed during the test case You can review the results of a line level analysis in the Source view This is the default level Note You need access to the source code for your application Line with Condition JProbe collects line level data plus it determines how many condition outcomes were exercised during the test case Some of the conditions tracked by JProbe include eif else edo while efor and e switch case default e assert statements Filters for Coverage The Analysis Type gt Filters tab is where you can set filters for your Coverage analysis Filters help you to focus on those aspects of the program that are important to you Your Data Collection Filters table may already contain a set of default filters The default filters are based on the description of your code that you provided see the My Code tab in the Configuration wizard You can use these filters or edit them For example if you know one of your methods is 100 covered you may choose to add a filter to stop collecting data about it Consider adding filters whenever you have achieved the level of test case coverage you desire for a particular method 370 JProbe User Guide Summary Analysis Type Java Application My Code Advanced JProbe Options JVM Analysis Type Memory Analysis Identify and resolve Memory leaks and object cycling Performance Analysis Identify and re
334. n particular the root of the graph and the thread nodes Other nodes are method types including standard methods and initialization methods By default thread nodes are hidden and are not included in calculations It can be helpful to think of the Call Graph as a set of superimposed call trees For example if two methods call methodA then methodA appears as a single node with two parents If you remove the call tree for one parent this is called pruning methodA remains because it has another parent The following table lists the types of nodes found in the Call Graph Icon Type Description m Root Conceptual in nature the root node is the ancestor of all threads or call trees if the thread nodes are hidden Exploring Performance Analysis Views 299 Exploring the Methods View Icon Type Description bee Thread Conceptual in nature a thread node is the ancestor of the methods invoked by that thread pa Encapsulated Represents methods that you are not interested in analyzing based on the encapsulated data collection filter StockSelectServiet lt Class gt lt method gt Represents a method for which I roturnChart 1 421 9 A you collected data at either method level or line level as above lt Class gt lt init gt javac compiles your constructors into initialization methods as above lt Class gt lt clinit gt javac compiles your class initialization statements into lt clin
335. n in Java source suchas if a amp amp b becomes multiple condition opcodes in the compiled bytecode For more information on how compilers generate bytecode for the JVM refer to The Java Virtual Machine Specification published by Sun Microsystems In particular see the following chapters e Chapter 6 The Java Virtual Machine Instruction Set http java sun com docs books jvms second_edition html Instructions doc html e Chapter 7 Compiling for the Java Virtual Machine http java sun com docs books jvms second_edition html Compiling doc html JProbe Coverage uses condition opcodes to collect condition coverage data For each condition opcode as your program runs JProbe will track which of the two possible outcomes true or false have occurred and record this in the snapshot In the Source View of the snapshot for each line which has a conditional opcode the Conditional Data table displays the conditional opcode mnemonics from the line and for each indicates whether execution of the condition resulted in a true or false outcome Note Opcodes generated by a compiler often differ from the source code You may see a condition in your source code that does not evaluate as you expect For more information see Nonintuitive Results for Conditions on page 396 388 JProbe User Guide The following table lists the opcodes that may appear in the Condition Detail table
336. n you indicate JProbe User Guide Task Export the chart to CSV Action Right click in the chart and select Export to CSV In the Export to CSV dialog box specify the export parameters and a location to save the file Response The file is saved to the location you indicate Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted Export the table to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Exploring Performance An
337. n your program enters or exits a method of your choice With advanced trigger controls you can also specify when to start executing the trigger actions and how long to execute them Triggers are useful for automated sessions Triggers are defined in the same way for all the analysis tools however the trigger actions are somewhat different This section describes the trigger actions for a Performance analysis To learn how to create triggers see Setting Triggers on page 425 Trigger Action Description Recording Starts recording data You can select an encapsulation option and the method profiling in the Recording Parameter area e Data Recording Off e Public Methods No Encapsulation e No Encapsulation e Grouped Encapsulation e Full Encapsulation Note When the trigger action is set to Recording the selected recording parameter is also displayed in the triggers table Snapshot Takes a snapshot You can provide a base name for the snapshot otherwise the snapshot file name takes the form snapshot_ lt n gt _raw jpp 272 JProbe User Guide 12 Exploring Performance Analysis Views This chapter provides detailed information about the Performance views used to present performance data This chapter contains the following sections Exploring the Runtime Summary ViGW cccsssseesesetesteetesnsetesneesteneesteseestesneetesneetenseateneeanens 274 Exploring the Summa VEN ngrira ieai i AENEA EETA 283 Exp
338. ncel Help l Refresh In the list of JVMs you see all active JVMs including the one that is running the JProbe Console on select platforms Select the JVM that is running the application and click OK JProbe connects to JVM With the exceptions of the Set Recording Level feature and the Heap Data tab which are not available you can interact with the session in the same way as in any other Memory analysis session such as requesting garbage collections and taking snapshots When you have your snapshots disconnect from the JVM by clicking Detach From Session F on the toolbar Examine the snapshots using the Memory analysis views Connecting JProbe to a Remote Application Started Without JProbe The following instructions apply if the program you want to attach to is running on another computer than the one running the JProbe Console 98 JProbe User Guide To connect a console to a remote application started without JProbe 1 On the remote machine from a shell or command window run the command jpattach with no options Note For more information about the jpattach command see the JProbe Reference Guide A list of Java applications to which you can attach to is displayed in the command window as shown in the following example JProbe version 8 3 0 Copyright 2009 Quest Software Inc ALL RIGHTS RESERVED JProbe is a registered trademark of Quest Software Inc Available target JVMs shown a
339. nd allocated objects are held by this node This node also includes native methods those that are implemented in some other language than Java For more information see How Data is Affected by Using Performance Features In this case you can select a method right click and select View Method Names to show the methods that the parent may have called and that may be included in this node Call Tree The Call Tree presents the same data as the Call Graph in a traditional more detailed tree view While the graph displays only one node per method the tree view has a node for every time a method is called and displays the nodes in the order of the calls weicome B Runtime Summary Summary Za Methods Nowe v Exploring Performance Analysis Views Exploring the Methods View Color to Show Hot Methods Bapsed Time fa me GF Cats PU Method Eo Method Y OPUCoi Elan Cmi Ava CPU Method o CPU Method Avg Enp Method a Elo Method croot E domos performance Aners Phiosop E demos performance diners Philo F java awt Graphics crawimaged Y ara lang mam singaouiey y java lang Mam co java ang Mam m EY jaran swing Imag T java awt Ora HRocsint ang i F java awt Graphics setColor ava a E demos perlormance diners Philo 3 E demos performance Aners Diners Cal Tree Cal Graph Each row in the Call Tree represents a method called by your program during its o 0 eecese
340. ne contains the Method List 298 JProbe User Guide welcome G Runtime Summary G Summary A Metnogs 7 me WR wl j pai E LM EELS J Cal Tree Cot Graph Fiter Methods 40 40 Cals CPU Method Clap Method Y CPuCmi Gap cat Avg FuMethod aCFumethod Avg Elap Method a Clap Mathod s 6 3 560 IM 5 1 853 1 866 1453 5 Prosi 0 8 Phitosopher rund 1 1 044 1 066 104 o E domos performance diners Pratosopher rung 1 600 1 865 1840 o E demos performance diners Prtosopher rund 1 1 607 1 067 1397 o E domos performance diners Pritosopher rund 1 631 1 285 18st o s T java lang Thread sieepdong 10 2 123 2 5 a W java tang Thread stay 2 2 0 0 ai E java io PrintStream printngava lang Stang 10 1 1 0 o DE cemos portoemance diners Priosopher painga awt 15 o 0 o si T java lang Thread init Qava lang Runnable java tang 5 o o 0 0 a E demos performance diners PrillosopherPanel drawPh s o o o o a RY arar swing Smngi tabes invokeLaterGars lang Runn 5 o 0 ox Recorded Levet Ful Encapadanion G oec a Methed C Apyeote hy Encaonisted Call Graph The Call Graph starts from the node provided to this view by the Performance Summary view The Call Graph is a directed graph that shows the calling relationships among executed methods in your program The nodes represent the methods and the edges represent the call from one method to another Some nodes are conceptual in nature i
341. neaneesees 85 NBOSS V4 2 accra anand nea andi nl E ETE E E E ay 86 JBOSS V5 Xipre aen aE N NE ATEEN GREER NE NGE 87 Open Source GlassFish V2 UR2 nenir eiii E EEE EEE EEEE E 87 OpenSource Glas FiS V3 a a a aa aan 88 Open Source Java Service Wrapper V3 3 cssessssscsssesssecesssecstssesstecsessessssenteeneestestesatecstesueestesseeateatesstesseensesass 89 Oracle AS 10g v10 1 3 0 or below Oracle AS 10g v10 1 3 1 or above Oracle WebLogic V10 3 s iccisedisiaeaiidinn aN a e E AE A N N 92 Oracle WebLogic Server 11g R1 v10 3 1 erssisavcneineunniineene aii Anca ate ananaeatan 92 OracleOC4 S109 VINO Kriisan aE raap AES EEOAE EEE a LERE 93 Oracle OC4J 10g v1 0 1 2 of below nc A ten oceania 93 Oracle OC4J 100 V1 01 3 vie etd beswaiie E A E ivadupan anaes 93 SpringSource dm Server v1 0 2 SpringSource dM Server V2 0 nimua aiao ehh le dina ERN aes aia 94 What to Do When the Console Does Not Connect c seessessessessseessestesseeseestesseestesseesneeseesneestsentesnsesneeteenteenseaneeneensees 95 Connecting a JProbe Console to an Application Started Without UProb sceseessesseesseseeseeseesesneestesstestesseseenesees 96 Connecting JProbe to a Local Application Started Without JProbe c csessessessessesssssessessessesstssessesenseesseseneaeess 96 Connecting JProbe to a Remote Application Started Without JProbe 00 eseessecseesteesteestesneesteestesneesteestesseeseesnee 97 Strategies for Memory Analysis s scssssssss
342. new settings file and click Save 70 JProbe User Guide The setting JPL file is created This is the JPL file that you use in the following procedure To modify the startup script to run an application under JProbe 1 Verify that you can run your application server with a JProbe supported JVM For more information see the list of supported platforms in the JProbe Installation Guide Make a copy of the application server startup script In the copy find java and replace it with jplauncher jp_input lt JPL_file gt jp_java lt path_to_original_java_executable gt For example in WebLogic you would find SJAVA_HOME bin java Xms64m Xmx64m classpath and replace it with SJPROBE_HOME bin jplauncher jp_input lt JPL_file gt jp_java sJAVA_HOME bin java Xms64m Xmx64m classpath where lt JPL_file gt is the fully qualified path to a JProbe settings file Use the quoting conventions required by the application server and or your operating system You can choose to specify the java executable in the JPL file instead of on the command line If you do not specify the java executable the java executable in the PATH environment settings is used You may also need to set JPROBE_HOME or use the fully qualified path instead Ensure that you have added all jar files to the classpath Save the modified script To run a JProbe session execute the modified script Starting Your Application with JProb
343. nfiguration settings and a JProbe startup script to launch your application Press Save amp CLose to create only JPL file contains the configuration settings Save amp Integrate Save amp Close Cancel Hep If a setting is incorrect you can go back now and change it by clicking the Edit link beside the setting You can also edit your configuration after the wizard ends 50 JProbe User Guide If all settings are correct you can save the configuration settings and also integrate your application to run with JProbe Tip The Save amp Integrate option is the recommended choice for first time users Saving the Configuration When you click Save amp Close JProbe saves the configuration settings To save the configuration settings 1 Inthe Save JProbe Settings File dialog box select a target directory for your settings file 2 If necessary rename the settings file The default file name contains the name of this configuration the type of analysis and the word settings The file extension is JPL 3 Click Save The new configuration file is saved in the directory of your choice and the new configuration name is added to the Configurations Navigator The lt Configuration_File_Name jpl gt tab appears in the View area From here you can e Further edit your configuration settings if necessary Click Edit The lt Configuration_File_Name jpl gt tab changes to edit mode For more
344. ng to the selected setting then guides you through the appropriate configuration steps Caution You may need to define some additional settings in the subsequent wizard screens for example filters and elements Configuring JProbe to Analyze Your Application 27 Configuring a Java SE Application e Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 e Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 Configuring a Java SE Application A basic setup takes only a few minutes though it may take longer if your application is complex or if you want to read more about the available options Note To integrate an unsupported application server follow the procedure in section Creating a Configuration for an Unsupported Application Server on page 65 Before you begin you should have the following information e The path to the main class of your application e The path to the supported JVM you want to use e Any arguments you use on the command line e The path to all class libraries required by your application if not specified in your default class path e The names of packages used in your application e The type of analysis you want to run Performance Memory or Coverage
345. nge the filter on the Derived class to method level the performance data will be correctly attributed to Derived method Troubleshooting Performance Analysis Results 351 Performance Session Runtime Tips The session hangs when running an application server under load If you notice excessive heap growth without garbage collections followed by an application crash increase the Xms Xmx parameters or lighten the load so that the JVM has a chance to catch up and do a garbage collection As the JVM caches data you can increase the load For example the compiling process for JSPs takes up a significant amount of memory If your session is failing at a load of only say five pages per second it is likely a memory allocation problem Running through the JSPs at a slow pace allows the application server to cache them and afterwards you can increase the load on the system without difficulty Active methods may report as lt unknown gt lt unknown gt after re starting data collection When not recording Performance does not keep track of where your program is executing When you re start recording it will track future calls but it may still not know what was executing when the recording started until it exits that method If that method is a long running method for example a run method on a Runnable the method may still be unknown at the time of the snapshot To avoid this situation start recording before the long running me
346. node Response The node disappears If the node had children that were not called by any other method the children are hidden Show a hidden Do one of the following The node is displayed as method e If you just hid the node click well as any other Back to go back to the previously hidden nodes previous graph that are required to link e Click the on the parent the DEW node to the node existing graph Clean up the display Right click the graph and select The nodes are rearranged of the nodes in the graph Navigating Subgraphs Layout Graph to improve readability While investigating hotspots you usually need to show more than the top methods For example you may want to show the top three methods called by a hotspot method to see which child method contributes the most to the method s cumulative time In other cases you may need to show an entire subgraph to find the real cause of a hotspot You can also hide subgraphs after you are sure the methods are not contributing to your performance bottleneck Hiding subgraphs reduces the complexity of the graph making it easier for you to follow the path you are interested in analyzing Use the child navigator to hide and show the methods that are called by a method The navigators are illustrated in the diagram below 322 JProbe User Guide Parent Navigator Child Navigator DO StockSelectSerlet E returnChart 1 421 9 The
347. node with dull child nodes the bright node is the hotspot 248 JProbe User Guide Using the Call Tree The Call Graph is useful for showing paths but because the nodes are large it is not very compact Sometimes the Call Tree is easier to navigate than the Call Graph The Call Tree displays the CCT see How JProbe Builds a Calling Context Tree on page 223 as a tree table that you can navigate by expanding and collapsing the nodes To use the Call Tree open the Methods view and click the Call Tree tab Investigating Performance Bottlenecks Now that you have identified a hotspot in your program you need to determine if it is really a candidate for improvement and if so what degree of control you have over the algorithm You can drill down to more information using the Method Detail view The Method Detail view displays details on the method in the call path for which you are currently viewing data For more information see Exploring the Method Detail View on page 325 To investigate a performance bottleneck 1 In the Call Graph select the hotspot method you identified The method is highlighted in the list and in the graph if visible 2 Right click the selected method and select Open Method Detail View The Method Detail view opens with the selected method displayed in the Current Method field 3 Evaluate how the Current method is called by reviewing the information in the Edges From Parents area 4 Evalu
348. not go to the Instance Detail view Using the Call Traces View The Call Traces view works almost the same way with garbage collection traces as it does with other recorded data see Exploring the Call Traces View on page 174 except that all the objects have been garbage collected so you cannot go to the Instance Detail view 122 JProbe User Guide Learning More About Memory Options This chapter presents how to fine tune the Memory analysis tool and how to analyze a heap dump This chapter contains the following sections Analysis Options for MOMOLY c sscsssseeseestessetesneetesnseesneeneneeaeeneesteaseatesnseesneeatenseaneneeanenses 124 moondng Neap DUMPE cs eaters oe cteneanes nus nea stevie tenia enenrane net tsieidian aoe eesnnirenmctin 127 124 JProbe User Guide Analysis Options for Memory You can configure the following data collection settings for your Memory analysis e Filters for Memory e Triggers for Memory e Initial Recording for Memory By default JProbe collects allocation data only You can collect stack traces and data on garbage collections by setting filters Filters for Memory You can select Data Collection Filters for your Memory analysis in the Initial Recording tab of the configuration wizard Unlike the other analysis tools you should avoid excluding code from a JProbe Memory analysis during early stages of your analysis If you do exclude code you may not see object allocations that
349. ns Detail be 5 OR Select a class right click and select the Show Allocations Detail option from the list The Allocations Detail view opens and displays all objects allocated directly or indirectly by the selected package class or method For more information see Exploring the Allocations Detail View on page 163 Note lf filters are used in the Allocation view the list contains objects that match the display filter View the source code for the selected allocation Select a class and click Show Allocated At Source OR Select a class right click and select the Show Allocated At Source option from the list Note If this menu item is disabled JProbe cannot associate a source file with the class file A browser opens and you can search for the source file of your application The source file opens at the code line defining the selected allocation For more information see Exploring the Allocations Detail View on page 163 Select how to display the allocation data Select one of the available options from the Group Data By list The table is updated to reflect the selected option 162 JProbe User Guide Filter the table Type a filtering expression in the Filter Allocation Points text box then press Enter The classes in the table reflect the filtering expression s you have entered For more information see Filtering Data in the Instances Vie
350. ns screen or afterwards in the Advanced JProbe Options tab Configuring JProbe to Analyze Your Application Specifying Advanced JProbe Options E New Configuration Wizard Advanced JProbe Options Specify Advanced JProbe Options Optional Specify advanced JProbe options For example you can set the absolute path to a directory in which to place Engine snapshots using jp_snapshot_dir lt directory gt Advanced JProbe Options R JProbe Port 52991 Snapshot Basename Save amp Integrate _ Save amp Close Cancel Help Advanced JProbe Options This field is for advanced JProbe command line options For most analyses you can 47 leave this field blank Command line options can be useful For example you can set an option to change the directory where the JProbe Analysis Engine creates snapshots For more information see the list of advanced JProbe options in the jplauncher section of the JProbe Reference Guide To specify JProbe options 1 Click the browse button beside the Advanced JProbe Options text box 2 Do one of the following e In the text box enter the options as you would on the command line and click Parse Arguments The table is populated with a list of options If an option was parsed incorrectly double click the cell and edit it e In the table double click an empty cell and enter an option Repeat for additional options 48 JProbe User Guide
351. nsole usually displays a message explaining why that action is currently disabled and suggests what you need to do to resolve the situation This section provides a list of memory help message you may encounter while doing a Performance Analysis using JProbe e No Line Data on page 353 No Line Data You can view line level details only if your snapshot contains line data To collect line data select the Line Level option when defining your data collection filters and run a new session to obtain a new snapshot For more information about line data see Metrics and Data Collection Filters on page 241 and Collecting Data on Lines of Code on page 255 If you have already set your data collection filters to collect line data and still get the No Line Data message it may be that the class was compiled without line information You may need to recompile the class with your compiler s debug option for example g in javac 354 JProbe User Guide 15 Strategies for Coverage Analysis The Coverage analysis tool provides a statement level assessment of your test case coverage It can also be configured to analyze how well your test cases cover the conditions in your source code You do not need a specific build a complete build or access to all the source code to start analyzing your test case coverage As long as you can compile the source code you can use the Coverage tool Tip JProbe does not analyze loops or c
352. nstances Click Show All Pinned Instances OE The lower pane is updated to display all instances pinned by the starting set whether immediately pinned or pinned indirectly through another instance See the instances that are pinned by a dominating instance Select a class right click and select the What Do These Pin option from the list The Pinned Instances view is updated to display the instances dominated pinned by the selected instance set Calculate Actual Keep Alive Size The sign in the Keep Alive Size column indicates an estimated value for this metric To calculate the actual size right click the row and select Calculate Actual Keep Alive Size from the list The Keep Alive Size column is updated with the actual Keep Alive Size value Save the chart as a PNG or JPEG file Right click in the chart and select Save Chart As then select the graphic file type PNG or JPEG in which you want to save the chart In the Export Image dialog box specify the export parameters and a location to save the file The file is saved to the location you indicate Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden Exploring Memory Analysis Views 159 Exploring the Pinned Instances View Resize column width
353. nt algorithm 10 Close the Snapshot Difference view Assessing Overall Program Performance When modifying an algorithm you often have to make changes to other parts of your program In this case you need to assess the trade offs of those modifications You may have improved the algorithm itself but added other code that impacts the runtime of the program Alternatively your algorithm itself may run more slowly but changes to code elsewhere improve overall program performance The overall program performance is the Cumulative Time for the entry point for your application For Java EE applications you need to find the entry point to your application In Servlets for example you would focus on the doGet and doPost methods For standalone applications you can look at the thread that runs your main method You cannot however see overall program performance for all threads at once To measure the program performance 1 Run the modified code using the original JPL settings file and take a performance snapshot For details see Running an Analysis Session on page 237 260 JProbe User Guide In the Snapshots Navigator load the performance snapshot that contains data on the original code For details see Loading Saved Snapshots on page 449 Right click the Performance snapshot containing data on the original code and select Snapshot Differencing The Performance Difference dialog appears In the Baseline Sn
354. nt to run Performance Memory or Coverage At the end of this process you will have a settings file and a startup script To create a configuration for an unsupported application server 1 Verify that you can run your application server with a JProbe supported JVM For more information see the list of supported platforms in the JProbe Installation Guide Launch the JProbe Console on the computer where you run your application server Windows Click Start gt Programs gt JProbe gt JProbe Console UNIX Navigate to JPROBE_HOME bin and execute jpconsole 2 Click Configurations in the Navigation area 66 JProbe User Guide Right click Java Application and click New Configuration or Select Java Application and click Configurations gt New Configuration Tip Alternatively you can use the standalone version of the Configuration tool called jjpcontig You can launch jpconfig from the Start gt Programs gt JProbe gt JProbe Config menu in Windows or from the JPROBE_HOME bin directory In a headless environment you can use the command line version of joconfig For detailed instructions see the JProbe Reference Guide The New Configuration Wizard dialog box appears Tip If your application server is not in the list you may still be able to use JProbe For more information see What Can Do if My Application Server Is Not Supported on page 64 Type in a name for this configuration and click Next For
355. ntifying the critical bottleneck code blocks with the least possible overhead If higher overhead memory detail was being captured at the same time the Performance results would be much less accurate e Coverage analysis is used in conjunction with unit testing and test suite generation This is often done by a test engineer or QA developer that is by someone other than the developer who wrote the code and ran Memory and Performance analyses with JProbe Coverage analysis is focused on determining which parts of the application have not been tested so Performance and Memory data is not relevant in this phase of the project Which Analysis Should Do First It depends on the types of problems you are seeing in your application If your application crashes you probably want to try a Memory analysis first If your application hangs run a deadlock detection within your Performance analysis If your application is slow our service consultants often recommend a Memory analysis followed by a Performance analysis The reasoning is that a memory leak will cause the server to crash eventually so you should address that issue first After the code is working and running in the context of the entire application a Performance analysis can help you to fine tune how your application runs A Coverage analysis can be done at any time Specifying Advanced JProbe Options You can specify these options during the wizard on the Specify Advanced JProbe Optio
356. ntime Summary View Garbage collection overhead is defined as the percentage of time the garbage collector is running compared to the overall elapsed time of your program A low number means garbage collection is having low impact on your program A high number can indicate a problem When you are attached to a running session JProbe shows the Garbage Collection Overhead below the Garbage Collections chart in the GC Data tab of the Runtime Summary view The table in the GC Data tab also shows a breakdown by garbage collector The names of the garbage collectors are provided by the JVM consult your vendor documentation for details JProbe shows each garbage collector the time it has used and the number of times it has been invoked Collecting Data for GC Tuning with the Runtime Summary View If you need to tune garbage collection you will need to consult documentation from your JVM vendor about how to do this This documentation will usually suggest using the verbose gc option to collect data for the tuning but this output can be very difficult to understand The Runtime Summary view shows similar data in a format that can help you understand the current garbage collection behavior and heap sizing to help you tune Start your application with JProbe and let it get to its normal operating state During startup the program behavior is different as many long lived objects are allocated for the first time and the garbage collector is adjustin
357. o investigate based on elapsed time data This table contains five suggested methods to Available only when determining hotspots investigate based on CPU based on CPU time time data Methods That Match This table contains five N A Session Filters to Investigate suggested methods that match the session filters to investigate based on elapsed time data This table contains five suggested methods that match the session filters to investigate based on CPU time data Available only when determining hotspots based on CPU time Each row in these tables contains a method Name and the Method Time which is defined as follows e When the Elapsed Time option is selected method elapsed time elapsed execution time of all threads combined 100 e When CPU Time option is selected method cpu time cpu execution time of all threads combined 100 Cells that contain percentage values also contain percentage bars The tables are not sortable and they always display only five rows Exploring Performance Analysis Views 289 Exploring the Summary View Double clicking a row opens the Methods view with the method in that row as the focus method Important If you select a new method to investigate when the Methods view is already open the Methods view is reloaded with the new method as the focus method All history on the Methods view is lost The What Next section guides you on how
358. o collect traces if you are investigating by Recorded Count Memory or garbage data if you are investigating by Dead Count Memory Click Specify Settings x to open the Set Recording Level dialog box then select the appropriate option to collect traces garbage collection data or both The stack traces are inaccurate in the Merged Allocation Points view For example consider the following Java program Method A line 100 calls Method B Method B line 200 Object o new Object Most JVMs do some optimization to the code and eventually can inline the method B into the method A After inlining the line number reported in the stack trace will be A 100 that is the line where A calls B not the line where Object o is created B is not visible in the stack trace You will know when you open the source code that inlining has occurred You should be able to determine which method was inlined from the code associated with the line number locate that method and investigate the problem If you are not familiar enough with your program to locate the method you can rerun the session without the JVM s optimizing compiler Your program will run slower but the accuracy of the allocation point information should improve because methods are not inlined For more information see Disabling the Just in time JIT or other Optimizing Compiler in the JProbe User Guide Also due to the way the JVM reports allocations the trace may be
359. o if omer thisThread 78 o D o break i 2 3 gotHovManyChopsticks 1 w a 1 SuingUes ities invokebater repaint u e 1 1192 Thread sleep sleepVatua 8 l o o if omer thisThread a o o break Lad i 3 2s System out printlninaze waits for Chopstick chl id s 61 31 025 31 025 synchronized chl w w 0 0 aft owner 1 thisThread o o 0 J break s s o o nuaieal 5 S 1 2 gotHowManyChopst icks 2 w s 1 3 SvingUtilities invokelater repaint 1 3 2 System out printininame acquires Cnopstick a lt gt Interacting with the Source View If you collected line information for some or all of the methods in this source code you can jump to a specific method in the code by selecting the method name from the Methods list You can also sort the columns of data in the right pane Task Response Show hide table Right click and select Show The selected columns are columns Hide Columns Select the displayed The rest are columns to show and click OK hidden Exploring Performance Analysis Views 337 Exploring the Performance Source View Task Resize column widths Action To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit Response The colu
360. obe to run your application see Configuring JProbe to Analyze Your Application on page 23 To create a configuration for a Coverage analysis from an existing configuration 1 Select the configuration in the Configurations Navigator 2 Right click the configuration name and select Copy settings for Coverage Analysis A copy of the original configuration appears in the Configurations Navigator The configuration settings are displayed in the Configuration Details tab 3 To edit the new configuration click Edit The Configuration Details tab is placed in editing mode Q Welcome x Configuration Details xI Run amp Connect P Run H Integrate Delete ef save As Editing MyApp_Coverage Configuration Name MyApp_Coverage Summary Analysis Type Java Application My Code Advanced JPrabe Options JVM Analysis Type Memory Analysis Identify and resolve Memory leaks and object cycling Performance Analysis Identify and resolve Bottlenecks and deadlocks Coverage Analysis Identify Unexecuted line of code P General JP Fiters JP Automation Trigger On Method Event Action Please enter a method to set a trigger on Aj Initial Recording at JVM Start O Data Recording Off Data Recording On Save Cancel a Type in a name for this new configuration for example MyApp_Coverage b Edit the settings on any of the tabs as necessary A basic Covera
361. ods that take a long time to execute and are the first place to look when you are tracking down performance bottlenecks Strategies for Performance Analysis 221 Understanding the Memory Pools Chart Understanding the Memory Pools Chart In JProbe Performance the Memory Pools chart acts as a progress bar for the performance of your program marking how long the program runs in real time It also provides an overview of how your program uses memory in the Java heap For example many tall spikes in the graph represent frequent large garbage collections In this case or if you are curious about memory in use for long periods of time you can use JProbe Memory to streamline garbage collection activities and improve the way in which your program manages object lifecycles welcome FH Runtime Summary Configure Pools Pools and GC Interval One Minute Classes Loaded 2007 Classes Unloaded 0 Memory Pools GC Data Threads Memory Pools Mb Memory Pools 4 eden space survivor space tenured Gen q Available 00 02 05 00 02 15 00 02 25 00 02 35 00 02 45 00 02 55 Memory Pool Distribution Used Mb Committed Mb Available Mb Maximum Mb Total 13 0 18 8 5 8 167 3 Heap Subtotal 1 9 Be 3 3 66 7 Eden Space 0 0 0 9 0 9 4 2 Survivor Space I 0 0 0 1 0 0 0 5 Tenured Gen I 1 8 4 2 24 62 0 Non Heap Subtotal 11 1 13 6 25 100 7 Code Cache I 1 0
362. of code Java SE Demo Configurations To nana demo select it in the Navigator and then Run amp Connect Network Memory Analysis Open demo settings Tutorial Diner Performance Analysis Open demo settings Tutorial Adventure Coverage Analyms Open demo settings Tutonal Java EE Demo Configurations To run a demo select it in the Navigator and then Run amp Connect jemory Analys s Open dano sctlngs Iutoral erformance Analysis Open demo scttmgs Tylorml k Coverage Amalyszs Open demo sethings Tutonal Doon show the Welcome pages when Frote starts You can shew them agan by selecting Help gt Show Welcome Pages Configurations Navigator The Configurations Navigator displays the list of configurations already defined for each of the Java applications and Java EE applications servers supported by JProbe You can also define new configurations modify existing ones and load saved configuration files 420 JProbe User Guide When the Configurations tab is selected on the Navigation area the View area displays the Welcome to JProbe Configurations tab which provides an overview of the JProbe configuration process It also includes a quick way to launch the New Configuration Wizard and allows you to define configuration settings for each JProbe analysis type The tab also contains links to the JProbe community page documentation and tutorials created for several demo applications shipped with the product
363. of encapsulation simplifies the information that is displayed in the Call Graph because the methods are not shown by name but rather by an asterisk and there may be fewer nodes Nodes marked with an asterisk appear in the Call Graph but their information does not display in the list of methods below the Call Graph Running a Performance Analysis 253 Investigating Your Results Root Main aks V5 Encapsulation using Grouped option If you compare the grouped call tree with the Unfiltered Call Tree on page 251 you can notice that all of the children methods H in the grouped diagram are represented by an asterisk placeholder The placeholders represent all encapsulated children of the calling method For example in the diagram the left most asterisk encapsulated all of the method time for H2 H5 and H3 None Option The encapsulated type None does not display any encapsulated methods in the Call Graph Root Main v2 v3 Encapsulation using None option If you compare the None diagram with the Unfiltered Call Tree on page 251 you can notice in the None diagram that none of the H methods display in the Call Graph In fact when you view the diagram there is no indication that these methods were invoked during the session All of the attributes of the children methods are directly attributed to the calling method For example the time spen
364. ogic 28 6 7 60 3 63 Room 60 0 10 44 2 B Bathroom 100 0 1 100 0 7 Kitchen 100 0 1 100 0 7 Class Summary The Class Summary breaks down the results in the Package Summary by class For each method in a class you get the total number of lines of code in the method and the percentage of missed lines A QA analyst might generate this report before investigating missed code and after modifying test suites He can then use these reports to communicate before and after results to his manager 404 JProbe User Guide A Packages Name Classes Missed Total Classes Methods Missed Total Methods Lines Missed Total Lines _ Total 33 3 6 45 2 3 58 1 210 demos coverage adventure 33 3 6 45 2 31 58 1 210 Per class summary for package demos coverage adventure Name __ Methods Missed Total Methods __linesMissed TotalLines _ Total 45 2 31 58 1 210 Adventure 42 9 7 63 9 61 RoomList 20 0 5 41 4 2 Logic 28 6 7 60 3 63 Room 60 0 10 44 2 a E Bathroom 100 0 1 100 0 7 E Kitchen 100 0 i i 100 0 E z Per method summary for class Adventure Name Lines Missed Total Lines Total 63 9 61 lt init gt 0 0 8 lt init gt String 100 0 a initializeRoomList 25 0 8 play 50 0 2 main String 70 8 24 printUsage 100 0 2 ceanup 100 0 8 Per method summary for class RoomList Name at Lines Missed Total Lines T Total 41 4 29 lt int gt InputStream 0
365. ollected so far you can start and finish a use case and take a snapshot Each analysis type can have one open snapshot If a snapshot is open and you run another session under the same analysis type the runtime view is for the current session but any other open views continue to display data for the open snapshot To display data from a different snapshot double click the snapshot entry in the Snapshots Navigator 22 Refining Data Collection and Presentation This chapter provides detailed information about JProbe filters triggers and categories This chapter contains the following sections Reining Data Gollection FIMES ccoccccsccessccecscshecesececnseteceveenaeesneasdeesecebeeaesaer sessetteredeedecsracte stared 422 OUING e ik asiinndannGd ddieid sek A de dGR a GRREAN eR nnGhaneddes 425 Creating and Managing Categories sessesssesecsesssetessseeeneesseseesteseestesnsetesneeateneeneneeanenes 432 422 JProbe User Guide Refining Data Collection Filters You can narrow your data collection by excluding some code from the analysis You can also expand your data analysis to include other custom packages or third party packages Bear in mind that the more data you track the slower the session runs and the more complicated the results will be This section contains the following information e Adding Filters e Changing Filter Execution Order e Removing Filters Adding Filters A filter is composed of two pieces of
366. ollowing options are available e Nanoseconds e Microseconds e Milliseconds the default e Seconds Time Precision Defines the precision used when calculating the time units displayed in Performance views The following options are available no decimal places or one two three four or five decimal places JDBC Truncation Style Defines the JDBC truncation method The following options are available e Truncate at the Beginning of Connection and Statement Strings e Truncate in the Middle of Connection and Statement Strings e Truncate at the End of Connection and Statement Strings default e Do Not Truncate Connection and Statement Strings For information about setting other preferences in JProbe see Setting Preferences on page 453 13 Reporting Performance Analysis Results You can generate reports based on some of the Performance views This chapter presents the Performance reports The steps for creating a report are described under Generating HTML Reports on page 440 This chapter contains the following sections Seting Report OPOS acct ate ireira ae aeaeaei a EEEN EATA TEES 344 Runtime Summan REPON scrii iii aiaa iaieiiea 344 EN a e E E E E nchtiteafi Gh doentaigtierted nei 345 o da e i AA A T 346 Momod Detail RODO eaaa aE 347 Perormance Diiferonce Raponsie aaa 347 344 JProbe User Guide Setting Report Options When you generate a Performance report the number of me
367. om the list The Call Traces view opens displaying the call tree and graph for the selected instances For more information see Exploring the Call Traces View on page 174 Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted Export table data to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 166 JProbe User Guide Create a report for Click Export Data to HTML The Create Allocations this view a Detail Report dialog box
368. omain name or a process For details see the help provided in the user interface below the fields E New Configuration Wizard BEA WebLogic v10 0 BEA WebLogic v10 0 Tell us about your WebLogic application server installation Server Startup Script C Program Files ApplicationServers BEA_Systems user_projects domains weblagic10 gamepack startWebLogic cmd J The usual startup script name is startWebLogic cmd or start lt DomainName gt Server cmd ex CAbealwiserver_1O samples domainswwi_serveribinistartvVebLogic cmd JProbe s Integrate process will not modify the chosen startup script A new startup script with modifications will be generated to start the WebLogic server with JProbe Note The modified script will closely mimic the selected startup script and it will also require the same command line parameters and environment variables settings that are needed for the chosen startup script Required Field lt lt Back For more information click the Help button or see Runtime Notes for Application Servers on page 75 If your application server or version is not listed you may still be able to use JProbe For more information see What Can I Do if My Application Server Is Not Supported on page 64 Specifying Your Java Application You can specify these options during the wizard on the Specify Your Java Application screen or afterwards in the Java Application tab JProbe User Guide i Ne
369. on Data csssscssssssssesssssssesssssessessessessesenssessessussessnseussessnsesteaneaeesteateaneateaneaneaneenes 120 Learning More About Memory Options ssscscsssssssssssesssssesesssesesateseeseesesnsenesseeenseaeeaseeensentenesteneeanentnanees 123 Analysis Options for Memory a a ca ce ies el cal aca it tara ata ad 124 Filters 10r Memory itscceycai secrsecereetsioerareiress sont aa a aV A T nr adnan arrears 124 Triggers for MOMOly siete ooa en N EAN eed SA ers 125 Initial Recording for Momo viiceAccasivercn os aetna nena an a a TAE 126 Importing Heap DUMPS sir icssiccesssci tases nisssteiessnaza asin edscsthacetbansateanaeaatiatatesiaccet sis de eaaecanasin narra anaes 127 Importing an IBM SVM Heap DUMP uu eseesessessesseesessteseeseesteseesteetesesensetseaseneeatsneeateaseatenteateaneateetsaeestenteanentennes 127 Importing an HPROF Heap DUMP iirst riani aia taaah aAa R EA Eaa ANa A EESOSA 129 Exploring Memory Analysis VieWS sssssssssssssessessssseserseeesersessssersessserseeseserseeaeeesseesseerseeaeeesseeeeseeaseesaneas 131 Exploring the Memory Runtime Summary VIEW s ssessessscsessessecsessecsessessessessesseseussessueesesaneaeeatsateaeeensaneateateateateesees 132 Understanding the Memory Runtime Summary View c cccsecsessessessnssessessessesseesesseesesseesneaeesnseneanesssenseeseneenes 132 Interacting with the Memory Runtime Summary View 00 csesseseeseseseestesseesteeseesneesesneeeneeseentesneenieeneen
370. on Description Q Start Recording Data Starts recording Coverage data A Stop Recording Data Stops recording Coverage data and takes a snapshot xaj Take a Snapshot Takes a snapshot Exploring Coverage Analysis Views 375 Exploring the Coverage Session Understanding the Data in the Coverage Session Data in the Coverage Session is presented in terms of hits that is the classes methods and lines of code that were exercised by your test case The Coverage Session has two tables The upper table displays a list of classes hit by the test case and shows the percentage of methods that were hit within each class When you click a class the lower table shows the names of the methods in the class whether or not the method was hit and the percentage of lines that were hit Note Data about condition coverage collected when Line with Condition is chosen in the configuration is not included in the Coverage Session view Condition data is saved to the snapshots taken during the session welcome FH Coverage Session Bj Snapshot Browser ge adventure Room demos coverage adventure Adventure pgp mm demos coverage adventure Logic 7 demos coverage adventure RoomList 0 Show Only Methods with Hit Lines lt 10 Method Method Hit e Hit Lines demos coverage adventure Raam checkSelecti Yes 100 demos coverage adventure Room setAssaciate No 0 0 demos
371. on page 186 e Interacting with the Heap Graph on page 187 Opening the Heap Graph The Heap Graph provides a visual map of the referrers and references of a selected instance Heap Graph view is available from the Instance Detail view To open the Heap Graph 1 Select an instance in the Instance Detail view 2 Click Heap Graph Daa The Heap Graph opens The upper pane displays a graph of the selected instance and its immediate referrers and references and the lower pane displays them in a table 186 JProbe User Guide welcome amp Runtime Summary i instances i l Instance Detail amp Heap Graph z Color to Show Interesting Unknown Uninteresting j tg rN E e gt Sim 0x3067 gt gt Object 0xd68 Sim Result 0x819 S E SS ES SS lt Name Referrers References Recorded Size Keep Alive Size java lang Object 0xd68 1 1 001 true 5 136 21 136 demos memory sim SimgResult 0x819 1 2 true 16 16 lt class gt demos memory sim Sim Result 5 004 4 false 344 344 demos memory sim Sim 0x3067 5 007 4 false 32 106 400 Understanding the Heap Graph The Heap Graph initially shows only instances or classes that are directly connected to the selected instance but you can expand the graph in the following ways e Double click the arrowheads gt to the left or right of a node e Select a node right click and select Show Referrers o
372. oncurrent licensing from the command line 1 2 From a command prompt navigate to the JPROBE_HOME bin directory To test that the license server is available enter e Windows jplicense test lt license_server gt lt port gt lt application gt e UNIX jplicense test lt license_server gt lt port gt lt application gt 472 JProbe User Guide 4 5 where license_server is the name of the server running the Quest Software License Server port is the port number that the license server listens on and application is the licensing application name For example jplicense test serve_lics 8133 licensing To set the license server use the following syntax Windows jplicense server lt license_server gt lt port gt lt application gt UNIX jplicense server lt license_server gt lt port gt lt application gt where license_server is the host running the Quest Software License Server port is the port number that the license server listens on and application is the licensing application name Repeat on each of the developer workstations You can run JProbe if a concurrent license is available Enterprise Licensing Enterprise licensing allows your entire development team to use JProbe This option is not covered in this guide For more information contact your Sales representative Index A about JProbe 16 about Quest Software 20 adding categories 432 folders 436 triggers 425 advanced trigger contro
373. only this method plus any methods called by this method Focusing on a method is particularly useful for Java EE applications where you want to focus on your application or a small part of a large application For example you might focus on a doGet or doPut method For more information see Isolating Subtrees on page 319 Removing Threads and Method Subtrees You may decide while investigating a snapshot that you no longer want to include the data for a particular thread or method For example you may decide that there is nothing you can do about the performance of the Java class loader and you want to assess the performance of the program without this method and the methods it calls You can select the java lang ClassLoader method and prune its subtree Other typical data that you may want to prune from the data set includes time spent waiting calls to Object wait and I O time calls to methods in java io For more information see Isolating Subtrees on page 319 Encapsulating Data on Methods Encapsulation means to hide details behind a well defined interface When one of the encapsulation options is used Performance tracks the call into the method and then just accumulates data until the method calls out to a non encapsulated method The encapsulated method is a black box hiding details about calls to other encapsulated methods In general encapsulation improves session runtime performance and reduces the
374. onsider using the g compiler option when you compile your program Improving Line level Data with the g Compiler Option JProbe offers line level analysis with its Performance analysis tool and its Coverage analysis tool To report on line level data JProbe reads from the line number table created when you compiled your code The line number table matches a range of bytes in the compiled bytecode to a line in your source code Depending on your compiler the match can sometimes be imprecise You can view your existing line and local variables tables by running javap 1 letter 1 with your compiled class file If you want to try to improve the quality of line information you can compile your code with the g compiler option The g option generates debugging information on line numbers and local variables and identifies the source file names 416 JProbe User Guide Learning How JProbe Works JProbe has four main components e JProbe Analysis Engine collects data on your application e JProbe Connection Manager manages communications e JProbe Console presents data in a graphical interface e JProbe Snapshots contains the data collected about your application Application t JProbe a Connection JProbe Analysis Engine Manager Console E JProbe Snapshot E JProbe Snapshot JProbe running on a single computer local session Learning How to Use JProbe Effectively 417 Learning How JProbe Wor
375. ooting Tips Memory Session Runtime Tips Memory Help Messages nirinsmarsaonrsnie nara N A aia NaN No Heap Dal siri E A eee No Stack Trace Data 2 3 vas ennai an a A AEN N a a a Operation N t V lid for Categories iisiin nennen eaaa AE ERE E IE RAER 216 Strategies for Performance Analysis ss sesssssesssnsununnnnunuuuunnonnunununnenunuununnonununnunnonnnnnnnnnonunnnnennunnnnnnrnnnnnnnnnnnns 217 Strategies for Achieving Program Efficiency seeeseseseseesesseeseesesnsesneeseeneesueesesneestesieesiessesessueseesneenenneenneens 218 Analyzing an Existing Program ea aeaa dR E AEAEE SENERE SAE erated SERE anion pa a aA SERES 218 Focusing on a Known Inefficient Algorithm csccscecsesessecsesesseseesesseseeeessesesseeneeesseeeeseesteesseeasenseteasineeseneateeseneaees 220 Understanding the Memory Pools Chatt scessessesessecssscnssstecnsssnesssesneeasecseesseesecseesueeseesueesueeaseeneeateenseaeenseaneeseeenees 221 Table of Contents 9 Understanding Performance Data iccviieasisiivadaaviniia delice ainanaita E 222 Collecting Data on Methods 2 sccccaeesiiichisswsinarn cher an anniin E R i wane 222 Building a Calling Context Tree cces tt ccsiveecshi settee oi Rea ee testes eae aac ti a ete de 222 How Data is Affected by Using Performance Features csssssssseseseesersesssneseesseessesasestssteansanesseateeneateaeers 225 Recursive Methods armonian a Moa ae cad ol ean a iene 232 Running a Performance
376. opcodes definitions 388 opening snapshots 449 opening the Allocations Detail view 163 Allocations view 160 Call Traces view 175 Coverage Source view 384 Dominating Instances view 148 Heap Graph view 185 Instance Detail view 166 Instances view 138 Leak Doctor view 190 Memory Difference view 198 Memory Source view 197 Merged Allocation Points view 181 Method Detail view 326 Performance Difference view 338 Performance Methods view 297 Performance Source view 335 Performance Summary view 283 Pinned Instances view 155 options for Performance analysis 266 JProbe 46 Oracle AS 10g 90 91 OC4J 10g 93 WebLogic 92 over allocations finding 117 reducing 119 P parent navigator 247 PDF generating 441 per seat license 468 percentages hiding bar charts 391 in the Heap Difference View 200 in the Snapshot Difference view 339 Performance analysis options 266 strategies 217 analyzing a program 218 assessing program 259 automation 268 changing the gradation of the color scale 323 collecting data on methods 222 cumulative time and recursion 233 customizing the Call Graph 323 data understanding 222 Index 481 Difference view 338 encapsulation 225 269 features how data is affected by using 225 filtering 225 267 full encapsulation 226 GC Data 276 grouped encapsulation 227 help messages 353 identifying deadlocks 260 performance bottlenecks 243 inefficient algorithms 220 investigating deadlocks 264 performance bottlenecks
377. or Application Servers e start Start Geronimo with JProbe in a separate window Note If you are running Geronimo 1 0 the port number is displayed in GERONIMO_OUT not in the command window Apache Geronimo v1 1 You run the JProbe generated startup script in the same way that you run your Geronimo startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Usage lt configuration_startup_script gt run start geronimo_args where e run Start Geronimo with JProbe in the current window e start Start Geronimo with JProbe in a separate window Note If you are running Geronimo 1 1 the port number is displayed in GERONIMO_OUT not in the command window Apache Geronimo v2 0 You run the JProbe generated startup script in the same way that you run your Geronimo startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JPro
378. ore generating this report sort the table by the metric in which you are interested 348 JProbe User Guide Methods by Method Time Avg Num Cumulative merhodii Pave n Method Num SE u Method i Time Cumulative Exception Calls Time ms z Time Catches 5 ms Time ms ms Exits demos performance polynomial Polynomial main ii Gavalang Stringt 1 0 2 982 o 1 928 o o o lt ROOT gt lt ROOT gt o 2 982 0 0 0 0 0 demos performance polynomial Polynomial evaluate int int double double o 2 982 3 101 1 928 2 106 o o 14 Troubleshooting Performance Analysis Results As you work with JProbe you may need help to get the most out of your analysis sessions This chapter presents a few tips for solving some Performance analysis problems and for running sessions efficiently For Performance analysis issues that are not covered here refer to Quest s technical support This chapter contains the following sections Performance Troubleshooting Tips s cssccsssssscesssesseessesnssessesnsasssesneaeesesnesneaeeesneneenesneneeaeans 350 Performance Session Runtime Tips csceccsesseestesteeteseetesneeeesneesteneeateseestesneetesneeinsnseaneneentens 351 Poriormante Holp Massages sinisiin aaa anA 353 350 JProbe User Guide Performance Troubleshooting Tips The following tips can help you with problems you may encounter See the ReleaseNotes htm file for other product related issues I am having difficulties
379. organization has subscribed with this license number you may not be able to license JProbe on your computer e When prompted enter the host name of the computer running your application e Ifyou want to run remote sessions add the host name of the second computer from which you want to monitor JProbe sessions and review the collected data Tip To find the host name of a computer enter hostname in a command prompt or xterm session 4 After you complete the instructions on the Web site your JProbe license file is mailed to the subscribed person s email address Save the license file to your computer 5 Now you can register the node locked license See Registering a License During Installation or from Within JProbe on page 466 Registering Node Locked Licenses You can register a node locked license while installing JProbe or you can skip this step and register a license when you launch JProbe Alternatively you can run a command line tool to register JProbe Registering a License During Installation or from Within JProbe This procedure walks you through registering a license file on the computer or computers named in the license file It assumes that the computer has a graphical display If not a command line utility is also available For more information see Registering Licenses from the Command Line on page 468 To register node locked licenses 1 On the computer running your application and the JProbe Analys
380. ort 2 Select one of the following options e Binary HPROF File e Text HPROF File 3 From the Select Binary Text HPROF File dialog box browse to the file that you want to import 4 Click Open The dump is converted to a Memory snapshot 5 Select and view the Memory snapshot as usual How to generate a heap dump from a Sun JVM Some Sun JVMs Sun Java 6 JVMs and the latest releases of 1 4 2 and Java 5 JVMs can be configured to generate a HPROF format heap dump when an out of memory exception is thrown To generate a dump automatically add the XX HeapDumpOnOutOfMemoryError option to the java command line when you launch your program If your program encounters an OutOfMemoryError the JVM attempts to write an HPROF format heap dump to a file named java_pidNNNN hprof in the current directory where NNNN is the ID of the java process You can import this file into JProbe You can also use the jmap tool with the dump live format b option to produce a dump See the jmap documentation for details 130 JProbe User Guide 6 Exploring Memory Analysis Views This chapter provides detailed information about the Memory views used to present memory data This chapter contains the following sections Exploring the Memory Runtime Summary View scsseseescesessesseeteseestesteetesneetesneeseeneeateneeane 132 Exploring the INSTANCES VIM oo fes ts cases asain tiseessestean ten nteenencsennincenenntin ieee 138 Exploring t
381. ou are interested in Instance Detail by Referrers Name Size Allocation Time Referrers References Allocated At Javax swing JButton 0X2513 448 00 05 974 3 25 LeakExample addButtonToPanel javax swing JButton OX2be6 448 00 06 846 3 25 LeakExample addButtonToPanel javax swing JButton OX2db2 448 00 07 102 3 25 LeakExample addButtonToPanel jJavax swing JButton OxX23ce 448 00 05 766 3 25 LeakExample addButtonToPanel javax swing JButton OX2a32 448 00 06 590 3 25 LeakExample addButtonToPanel javax swing Button 0X1f58 448 00 05 383 3 25 LeakExample addButtonToPanel javax swing JButton OX30c2 448 00 07 342 3 25 LeakExample addButtonToPanel javax swing JButton 0X2675 448 00 06 166 3 25 LeakExample addButtonToPanel javax swing JButton OX27ec 448 00 06 374 3 25 LeakExample addButtonToPanel javax swing JButton 0X22a4 448 00 05 583 3 25 LeakExample addButtonToPanel Call Traces Report This report lists the locations of all the instances in the Call Traces view in order of cumulative count If you are investigating by Dead Count Memory these are the locations of garbage collected instances otherwise they are the locations of live instances Before generating this report sort the table by the metric that you are interested in Garbage Collected Locations by Cumulative Memory Name Cumulative Cumulative Count Memory System Code 207 4 968 com jprobe engine bci NormalObjectEventHandler newObject java lang Object
382. ou use an application server startup script If the JProbe Analysis Engine is on a remote machine you can use the jpengines tool on that machine to find the port number that the Analysis Engine is using and to verify that it is actually running For details about the jpengines tool see the JProbe Reference Guide 96 JProbe User Guide Incorrect Host Name Click Attach to Session on the toolbar and ensure that the information specified in the Host Name IP Address matches the host name of the computer running your application under JProbe Application Not Started Under JProbe Make sure that you launched your application using the JProbe generated startup script rather than your usual startup script Connecting a JProbe Console to an Application Started Without JProbe This feature has the following restrictions e Requires Sun JVM version 1 6 or later e Works only with the JProbe Memory analysis tool e The ability to Set Recording Level and the Heap Data tab are not available but you can still request garbage collections and take snapshots If you are using a Sun JVM version 1 6 or later you can attach JProbe to the JVM while your application is running This may be useful in a pre production environment where you want to analyze a problem as it is happening For example your quality assurance team may be running a set of tests during which they notice that the application has slowed down over time You can attach JProbe to the
383. ounter overflows Quality Assurance analysts use the JProbe Coverage tool to assess how much an application s code is covered by their test case or suite of test cases Software developers can also use the Coverage tool to verify the completeness of their unit tests and to locate obsolete code This chapter provides recommandations for how to use the JProbe Coverage analysis tool in your development environment This chapter contains the following sections Assessing an Existing Test Suite s sssssssesrssesssssrrstensssnrrrtensssntnttennnntnttrennsnnrineennnnrinrernsnna 356 Determining Overall Coverage sessssssssssssisstesssssrrstensssnrrttensnnnrittennsnnrittennsnnnrntrnnsnnniitrerasanrnnet 356 Modifying a Test Suite 0 Creating a New Test Suite Keeping Up with Development c ccscssscsscesssssseesseenseceesnesteaeaneaeeaesneateaesesneaeeatsneaeeaesneaeeaeass 357 356 JProbe User Guide Assessing an Existing Test Suite If you have a large complex or long running process you can perform a preliminary analysis of your test suite at the method level The sessions run quickly and you can clearly see which methods are missed entirely by your test cases Once these areas have been identified and corrected you can then follow up with more thorough line by line analyses to find missed lines of code In some cases your method level analysis may point out blocks of code that were not exercised because your te
384. oup Threads Displays threads grouped thread nodes i or ungrouped in the graph and trees Restore the graph to Click Reset ni Resets the graph to its the before edits default display criteria state and clears history of operations All methods removed through pruning or isolating are restored top methods that were hidden are shown and all other nodes that were shown are hidden You cannot undo a reset action Exploring Performance Analysis Views 307 Exploring the Methods View Task Action Response Move around in a The Call Graph shows the large graph Click Show Graph Panner to open the Graph Panner window then move the rectangle to the part of the graph that you want to see section of the graph that you selected Create a report for this view Click Generate Report er to open the Create Methods Report dialog box Select a location for saving the report and click Generate File The report is generated in the specified location Undo an edit Click Back O The Method view reverts to the state before the edit took place You can undo reset isolate and prune actions Note You can not undo the reset action Redo the last undo Click Forward Reapplies the edit on the view Customize the fonts See Customizing Fonts on page 324 Changes the font family size and style of text in the graph 308 JProbe User Guide Task A
385. our application and the JProbe Analysis Engine do one of the following e To set the license server while installing JProbe select the Concurrent Licenses tab in the Licensing Setup installation screen Note This option is available only if you choose to run a Custom installation To set the license server after installation launch JProbe and select the Concurrent Licenses tab in the JProbe Licensing dialog box Type the fully qualified host name or IP address of the computer running the License Server in the License Server Hostname field Click Test Server A message tells you whether or not the computer is available To set the specified host as your license server click Set Server To view your license and the licenses currently in use click View License Usage e The License Server home page opens in your default browser e For details on each screen click the Help link e To close the License Server web interface close your browser To end the license registration process do one of the following e From the Licensing Setup installation click Next e From the JProbe Licensing dialog box click Close Repeat on each of the developer workstations You can run JProbe if a concurrent license available Setting the License Server from the Command Line You can test and set the license server from the command line For more information see Using Concurrent Licenses on page 469 To set the license server for c
386. performance problems you can be confident that any performance improvements you subsequently implement will be visible to your customers Tip The JVM can influence the performance of your program over time as optimizations are made You may want to run the program for a while and execute the use case a few times until you begin to get fairly consistent values and then take the snapshot Setting the Recording Level Before you start your analysis you need to decide what kind of data you want to see The Set Recording Level dialog box presents the following options Select this option In this situation Data Recording Off You are at the initial stage of your investigation when you do not know what are the performance bottlenecks This is often enough to at least start your investigation Public Methods No You collect data only for the public methods Encapsulation matching your filters No Encapsulation You collect data only for the methods matching your filters Grouped Encapsulation You collect data for your methods and combined data for the methods they call Running a Performance Analysis 237 Running an Analysis Session Select this option In this situation Full Encapsulation You collect data for your methods and the methods they call To set the recording level for your application move the slider up or down The selected option is highlighted in blue The position of the slider indicates t
387. plained within the main java file for the sample application The tutorials are available in a PDF called JProbe Tutorials If this PDF is not included in your installation you can download it from the JProbe Community site Learning How to Use JProbe Effectively 419 A Quick Tour of the Interface A Quick Tour of the Interface The JProbe Console has a main menu a toolbar a Navigation area left pane and a View area right pane The Navigation area contains two tabs Configurations and Snapshots They are referred to as Configurations Navigator and Snapshots Navigator respectively The View area displays different information depending on the tab selected on the Navigation area E Probe Console 15 Contepronioes IBY Snapshots A wekome Configuration Details pahe noarm Welcome to JProbe Configurations JProbe configurations define the process of configuring JProbe to run your Java EE and Java SE appbcabons To switch to analyze JProbe maprhot data select EM Snapshots tab at the top of the Configurations view Quick Start Tutorial amp Help Leam how to use JProbe by folowing a tutorial Tutorials are also available from help menu A New Memory Analysis Identity and resolve Memory beaks and object cycling JProbe Tutorials A Nen Performance Analysis a JProbe Community Identify and resolve Bottlenecks and deadlocks Probe User Guide A Nen Coverage Analysis Probe Reference Guide Identify Unexecuted lines
388. pmn xmil file to its original state and deletes the backup Caution Do not end this script using Ctrl C It leaves your application server in an unstable state Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly 92 JProbe User Guide Oracle WebLogic v10 3 You run the JProbe generated startup script in the same way that you run your WebLogic startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables and the same command line arguments Usage lt configuration_startup_script gt Note Windows XP and Windows Vista can be used to develop applications but are not supported by Oracle as production deployment platforms Oracle WebLogic Server 11g R1 v10 3 1 You run the JProbe generated startup script in the same way that you run your WebLogic startup script
389. ponse The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted Export the table to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Create a report for this view Click Export Data to HTML m The Create Merged Allocation Points Report dialog box appears For details see Merged Allocation Points Report on page 208 Exploring Memory Analysis Views 185 Exploring the Heap Graph View Exploring the Heap Graph View When you want to visualize the references among instances in the heap as a directed graph you can use the Heap Graph view Each node in the graph is an instance and the edges represent the references from one instance to another The table below the graph is similar to the one in the Instance Detail view it lists the instances in the graph the number of referrers to and references from it and its size This section contains the following information e Opening the Heap Graph on page 185 e Understanding the Heap Graph
390. pplication Running Your Script and Connecting the Console If during the integration stage see Executing the JProbe Startup Script on page 53 you have selected to run the JProbe startup script and connect the JProbe Console the script is launched automatically the JProbe Console is automatically attached to your running application and you can start to analyze your application right away If you want to specify any additional arguments to the script file or change the working directory you can use the Run button Note Before running the JProbe startup script you must manually set any environment variable required by your application You can also run and connect any JProbe script manually using the JProbe Console To manually run your startup script and connect the JProbe Console 1 Inthe JProbe Console select a configuration from the Configurations Navigator The lt Configuration_File_Name jpl gt tab appears in the View area 2 Click Run amp Connect The selected application server launches and the JProbe Console automatically connects to your application You can start analyzing your application using the Runtime view Configuring JProbe to Analyze Your Application 59 Managing Configurations Managing Configurations This section explains how to copy edit and delete configurations as well as how to manage settings files and review errors It also provides a list of pre defined demo configurations installed
391. pret the results of your JProbe analysis This section helps you prepare to use JProbe e Understanding the Architecture for Your Program e Selecting Use Cases e Compiling Your Program Understanding the Architecture for Your Program Review the architecture document for your program Architecture documents typically outline a high level view of the relationships within your Java program For example in a multitiered application the architecture would identify what each tier does how they communicate with each other and what data they exchange In addition a core set of principal classes methods and interfaces should be identified It is important that you understand the lifecycle of the principal objects when they are created how long they live and when they can be removed Focus on understanding the control flow that is how the objects within your program relate to one another If you do not have an architecture document consider creating one The Performance Summary and Methods views are a good place to start as they can give you an overview of the control flow in your program Selecting Use Cases Use cases describe high level tasks from a user s perspective If your development environment is process conscious look for documented use cases wherever project information is kept In less formal development environments you may need to derive the use cases by examining the program and end user documentation By focusing your
392. procedure you will have one or more Coverage snapshots By default a Coverage snapshot contains line level data on classes in your application Only classes that contain methods called during your test case are included in the snapshot Note This procedure assumes that you know how to start an application under JProbe If not see Starting Your Application with JProbe Attached on page 71 362 JProbe User Guide To run a basic Coverage analysis 1 Launch your application using the JProbe generated startup script Caution For application servers you may need to set environment variables and add command line arguments JProbe Execution Console demos memory leakexample LeakExample ol x Probe version 8 3 a opyright 2009 Quest Software Inc ALL RIGHTS RESERVED Probe is a registered trademark of Quest Software Inc lemory analysis started Probe Connection Manager is available for Console connections on port 52991 2 Launch the JProbe Console 3 Click Attach to Session f on the toolbar JProbe is searching for currently running engines and JVMs and displays this list in the Attach to Running Session dialog box Attach to Running Session Select the engine or enter the Host Name and Port Number below the table Name Product User com sun enterprise glassfish bootstrap ASMain Memory azafar Memory Memory lupas Host Name IP Address localhost Port 52991 v OK
393. propriate engine from the list and click OK to connect JProbe to your application The Runtime Summary view appears For a description of this view see Exploring the Memory Runtime Summary View on page 132 114 JProbe User Guide welcome x A Runtime summary zRo oT ke Pools and GC Interval One Minute Classes Loaded 1086 Classes Unloaded 0 na Memory Pools GC Data Heap Data Memory Pools Mb Memory Pools 6 eden space J survivor Space 4 tenured Gen Available is 0 00 01 45 00 01 55 00 02 05 00 02 15 00 02 25 00 02 35 Memory Pool Distribution Used Mb Committed Mb Available Mb Maximum Mb Total 7 0 18 5 11 5 167 3 Heap Subtotal 0 5 5 2 4 6 66 7 Eden Space I 0 1 0 9 0 8 4 2 Survivor Space 0 0 0 1 0 1 0 5 Tenured Gen 0 4 4 2 3 7 62 0 Non Heap Subtotal 6 5 133 6 9 100 7 Code Cache 0 7 0 8 0 0 33 6 Perm Gen a 5 7 12 6 6 9 67 1 Session Connected Finishing current recording Snapshot C Quest_Software JProbe_8 3 snapshot_raw_1_ jph has been added to the snapshot navigator as snapshot Finished recording snapshot C Quest_Software JProbe_8 3 snapshat_raw_1 jph successfully written 4 Before proceeding with the Memory analysis you can choose at this point to explore the data and ask a few questions such as is my application using about the right amount of memory If the application uses too much memory you could then i
394. ps are usually much larger than PHD dumps JProbe can import either format but with some versions of the JVM PHD dumps are missing information about array types Optionally you can set the IBM_HEAPDUMPDIR lt directory gt environment variable This environment variable controls the location where the heap dump file is saved Starting with SDK 1 4 2 some variables automatically set by the JVM IBM_HEAPDUMP_OUTOFMEMORY and IBM_JAVACORE_OUTOFMEMORY can result in several heap dump files heapdump txt and Javacore files javacore txt being generated when the Java heap becomes exhausted This can be disabled by setting the following environment variables IBM_HEAPDUMP_OUTOFMEMORY false IBM_JAVACORE_OUTOFMEMORY false Learning More About Memory Options 129 Importing Heap Dumps Importing an HPROF Heap Dump Most JDKs include the HPROF utility a simple command line tool for profiling your application s heap or CPU usage JProbe supports importing a HPROF heap dump in binary and text formats This may be useful if you use the jmap or jconsole tools in the JDK to produce heap dumps or if you use the HPROF utility for other purposes in your development environment and would like to be able to view the heap dump information that you collected in a graphical interface The heap dump is converted to a snapshot that you can analyze using the JProbe Memory views To import a binary HPROF dump 1 Select File gt Imp
395. pshot into the JProbe Console To load a snapshot 1 Click File gt Open Snapshot 2 Navigate to and select a snapshot file Tip In Windows each type of snapshot file has a specific icon associated with it 3 Click Open The snapshot is added to the list in the Snapshots Navigator Note If your snapshot is very large it is recommended to use the jpdumpreport tool to generate a report for this view before importing it into JProbe The jpdumpreport tool is an alternative for snapshots that are too big to be loaded but unlike jpprepare jpdumpreport does not alter the snapshot and it does not make it load any faster For details see the JProbe Reference Guide Unloading Snapshots Unloading a saved snapshot simply removes it from the Snapshots Navigator you can open it again at any time For example if you take a snapshot and then immediately take another that better represents the data you can avoid confusion by unloading the previous snapshot Unloading an unsaved snapshot prompts you to save the snapshot If you do not save it the snapshot is deleted To unload a snapshot 1 Inthe Snapshots Navigator right click a snapshot and select Unload Snapshot 2 If you chose a temporary snapshot you are prompted to save it e To save the snapshot click Yes and save the snapshot as usual e To delete the snapshot click No e To leave the dialog box without unloading the snapshot click Cancel The snapshot is removed from the
396. r If you have more than one kind of snapshot loaded you will have more than one current snapshot For example if you have both Coverage and Memory snapshots loaded you have two current snapshots one supplying data to the Memory views and one supplying data to the Coverage views To set the current snapshot 1 Inthe Snapshots Navigator double click the snapshot The snapshot name is changed to bold text to indicate that it is the snapshot providing data to the views A default view may be displayed 2 To open a view right click the snapshot and select Open Snapshot Saving Snapshots You can save snapshots to review later For example you can compare results before and after a change to the program When you save the snapshot rename the file to something other than the default snapshot name or save it in a different directory Files with the default name in the target snapshot directory are deleted when you exit JProbe unless you explicitly save them from the Save Snapshots dialog box For more information see Saving Snapshots on Exit on page 450 To save a snapshot 1 Inthe Snapshots Navigator right click a snapshot and select Save Snapshot As 2 Navigate to a directory of your choice 3 Type a name for the snapshot in the File Name field 4 Click Save The name of the snapshot is updated in the Snapshots Navigator Managing Snapshots 449 Loading Saved Snapshots Loading Saved Snapshots You can load a saved sna
397. r Show References When the graph expands beyond the size of the pane you can use the Panner to focus on various areas of the graph or use the scrollbars to move around the graph You can also use the Scale slider below the graph to enlarge or shrink the graph When you click a node in the graph the corresponding line in the table below is highlighted and vice versa The Instance Detail and Leak Doctor toolbar buttons then become available for further investigation of the instance The default status color scheme for the Heap Graph is Interesting Unknown Uninteresting You can select an alternate scheme Created When Recording Not Recording or Keep Alive Size from the Color to Show list above the graph You can hover over the color scale at the bottom of the graph to see what they represent You can change the status of a node by selecting it right clicking and selecting one of the options under Mark As or Mark Reachable As If you choose the latter the changed status will be applied to all the nodes below this one in the graph This can help you to keep track of instances that you have already investigated and have decided are definitely not involved in your leak Exploring Memory Analysis Views 187 Exploring the Heap Graph View Customizing the Heap Graph You can change the memory units displayed in the method list Tools gt Edit Options gt Data Display gt Memory You can also customize your Heap Graph view by changing the font
398. r use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables and the same command line arguments Usage lt configuration_startup_script gt BEA Systems WebLogic v10 0 You run the JProbe generated startup script in the same way that you run your WebLogic startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly 82 JProbe User Guide JProbe requires the same environment variables and the same command line arguments Usage lt configuration_startup_script gt IBM WebSphere AS Community Edition v2 0 0 1 You run the JProbe generated startup script in the same way that you run your WebSphere startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not wor
399. ration Tip Alternatively you can use the standalone version of the Configuration tool called jpconfig You can launch jpconfig from the Start gt Programs gt JProbe gt JProbe Config menu in Windows or from the JPROBE_HOME bin directory In a headless environment you can use the command line version of joconfig For detailed instructions see the JProbe Reference Guide The New Configuration Wizard dialog box appears Tip If your application server is not in the list you may still be able to use JProbe For more information see What Can Do if My Application Server Is Not Supported on page 64 Select a configuration type provide a name for this configuration and click Next For more information see Defining a JProbe Configuration on page 29 Enter the configuration data of your application server installation Select a supported JVM type any Java options that you use to run your application and click Next For more information see Specifying Your Java EE Application Server on page 31 26 JProbe User Guide 6 Specify application filters enable JProbe to create a category containing the code set to include and click Next For more information see Specifying Your Code on page 38 Select the type of analysis you want to run For more information see Specifying a JProbe Analysis Type on page 45 In this screen you have the following options e Continue the configuration proc
400. rawImage java awt Image int int java awt image ImageObserver 10 7 java lang Math random 3 6 demos performance diners Philosopher 1 run 3 6 java awt Graphics filRect int int int int 3 6 Methods That Match Session Filters to Investigate Method Method Elapsed Time of B demos performance diners Philosopher 1 run 3 6 demos performance diners Philosopher run 3 6 demos performance diners Philosopher paint java awt Graphics 3 6 demos performance diners Philosopher start int 0 0 demos performance diners Philosopher gotHowManyChopsticks int 0 0 When the JDBC information is present the Performance Summary report also includes the connections and SQL statement sections Methods to Investigate Method Method Time of A J java lang Class forName java lang String 38 1 java sql DriverManager getConnection java iang String java util Properties 0 2 java sql Connection setAutoCommit boolesn 01 java sql Statement execute java lang String 0 0 java sql PreparedStatement executeUpdate 0 0 Methods That Match Session Filters to Investigate Method Method Time of A SimpleApp goljava lang String 0 0 SimpleApp main java lang String 0 0 SimpleApp lt init gt 0 0 SimpleApp loadDriver 0 0 SimpleApp parseArguments java lang String 0 0 Connections by Elapsed Cumulative Time Method Calls Elapsed Cumulative Time ms jdbciderby derbyDB create true 1 1 390 jdbe derby shutdown
401. relative relationships among expensive nodes in the graph in particular when a graph has a few nodes with values that far exceed the other nodes There will be times however when you want a quantitative view of the graph In this case you can change the scale gradation to linear the color scale is divided into segments of equal size and the graph highlights the few most expensive nodes To select the gradation used for the scale e To highlight quantitative differences among the nodes right click the color scale and select Linear Mapping e To highlight relative relationships among the nodes right click the color scale and select Geometric Mapping default Changing the Color Scheme You can select from a few predefined color schemes In all schemes the bright color represents the most expensive methods while the dull color is for the least expensive method Select a color scheme that suits your preference If you want to print a graph in black and white the best scheme to use is Blue to Light Gray 178 JProbe User Guide To change the color scheme Right click the color scale and select a color scheme Customizing Fonts Use the Font Chooser dialog box to change the font family size and style of text in the graph To change fonts 1 On the toolbar click to open the toolbar menu and select Set Font The Font Chooser dialog box opens Font Chooser Name Style Dialog N Regular Sample
402. rence between Used and Committed memory Max The maximum amount of memory that pool can use Not all pools have a maximum value Maximum value is often affected or directly set by JVM options like xmx A Max value is always greater than the Committed value GC Data Tab The GC Data tab contains a graph and a table The graph displays each type of garbage collection at the time it occurred The Count scale units at the right of the graph indicate how many collections were done of each type The Garbage Collection Overhead data below the graph indicates the percentage of time the JVM was exclusively performing garbage collection compared to the elapsed time the program has been running For example a program that has been running for 10 seconds and has spent 0 5 seconds doing garbage collection has a GC overhead of 5 The lower the percentage the lower the overhead If the percentage is high you may need to tune your JVM s GC options or reduce object cycling The table provides a breakdown by garbage collector that shows the number of times each collector has been invoked and the time in milliseconds that it used as well as the cumulative counts and times Heap pools contain things normally found in the Java heap like Java objects Non heap pools normally contain things the JVM uses internally such as the compiled representation of method code or class information Threads Tab The Threads tab contains a graph and a list which are
403. rent Navigator Child Navigator NEEL returnChart 1 421 9 In many cases the hotspot will be represented as an encapsulated node If the real hotspot is from someone else s code you still need to focus on your own calling method Try to reduce calls to an expensive method or try to find more efficient ways to call it To identify a hotspot by navigating the Call Graph 1 Ensure that the Color to Show metric is set to show Hot methods Elapsed Time or CPU Time Adjust if necessary 2 Select the expensive method node in the Call Graph If the node is not visible right click the method and select Fit Method Into Graph to show it in the Call Graph 3 With the method node highlighted in the Call Graph click the child navigator Note In some cases a hotspot is a small frequently called method such as an accessor It is unlikely that this is your performance bottleneck but it may be hiding other hotspots 4 Ifall the expensive child methods are encapsulated the hotspot remains the calling method Usually they will be methods from third party packages and frameworks but if an expensive encapsulated node represents your own code you should rerun the session with a method level data collection filter set on the method 5 If one of the child methods is very bright and has method level detail select the child node Expand its children if it has any Repeat until you find a bright unencapsulated leaf node or a bright
404. resented by this node View Thread Names Opens the Threads dialog box which contains a list of thread names represented by this grouped node Exploring Performance Analysis Views 313 Exploring the Methods View Task View the list of methods represented by this node Action select a node in the Call Tree tab right click and select View Method Names Response Opens the Methods dialog box which contains a list of method names represented by this grouped node Drill down to source code Show Source If you have the source code the Source view opens with the source code displayed Drill down to line data Show Line Data The Source view associated with the selected method opens Disabled if line data is not available for the selected method Prune a subtree Prune Selected Node The selected node and its subtree are removed from the graph All metrics are recalculated without the contribution of the pruned call tree For more information see Pruning Subtrees on page 319 314 JProbe User Guide Task Isolate a subtree Action select a node in the Call Tree tab right click and select Isolate Selected Node Response The selected node is anchored to the root node All nodes that do not belong to the method s subtree are removed All metrics are recalculated to include only the data collected on the methods in the isolated
405. riable 3 To add another path required by your application click an empty cell and do one of the following e Click the cell browse button navigate to the path and click Open e Double click the cell and enter the fully qualified path 4 To add an environment variable double click an empty cell and enter the name of the environment variable using the form VAR or VAR either one works on both Windows and UNIX platforms Repeat for additional environment variables Note While not advisable you can choose to use your CLASSPATH environment variable To do so click Add Classpath Environment Variable 36 JProbe User Guide 5 To rearrange or delete paths and environment variables from the table click a row and use the buttons provided 6 To have JProbe propose default filters select the check box beside each path that you want JProbe to search for package names Classpath One Per Line C Program Files JPrabe 8 0 els 7 Click OK The classpath is listed in the Classpath text box Specifying a Java Virtual Machine You can specify these options during the wizard on the Select a Java Virtual Machine screen or afterwards in the JVM tab New Configuration Wizard J M Select a Java Virtual Machine Select a supported JVM and specify the Java options if any that you use to run your application Java Executable C Quest_Software JProbe_8 3 bin jre bin java exe Java Options
406. right side of the pane e The thread name and ID e The thread state BLOCKED and the object on which the thread is locked on e The number of times the blocked thread tried to enter or re enter a monitor that is the number of times the thread has gone into the BLOCKED state and the number of times the thread waited for notification that is the number of times the thread has gone into the WAITING or TIMED_WAITING state e The name and ID of the thread locking the selected thread e The stack trace and source code line of the method that caused the deadlock 11 Learning More About Performance Options This chapter presents how to fine tune the Performance analysis tool and how to analyze a heap dump This chapter contains the following sections Analysis Options for Performance cscssessteseeteseetesnsetesneeteeneesteseestesteatesneetesneeitenseneneeatenses 266 266 JProbe User Guide Analysis Options for Performance You can configure the following data collection settings for your Performance analysis e Timing Settings for Performance e JDBC Component Instrumentation e Filters for Performance e Automation Settings for Performance e Triggers for Performance Analysis Type Memory Analysis Identify and resolve Memory leaks and object cycling Coverage Analysis Identify Unexecuted line of code A General Filters Automation Timing Q Elapsed Ocu JDBC Component Instrumentation
407. rmance Hotspots section suggests methods to investigate based on Elapsed Time data or Elapsed Time and CPU Time data when available In the latter case you can switch between the two types of data by selecting the CPU Time or Elapsed Time option button respectively Note The CPU Time and Elapsed Time option buttons are displayed only when CPU data is available The CPU data is collected if the CPU timing option is selected when defining the application settings for details see Timing Settings for Performance on page 266 The hotspots are determined by searching for the top five most expensive methods by method time The upper hotspots table contains the top five methods from all of the methods for which we have data available Tip If the recording level is set to grouped grouped methods are omitted from this table as they are not a good starting point for investigation The lower hotspots table contains the top five methods that match the session filters defined by users when collecting data Therefore this table contains only non encapsulated and non grouped methods that match the filters set in the Create Edit settings dialog on the Analysis Type gt Data Collection Filters tab 288 JProbe User Guide The following table lists the type of information displayed in the Performance Hotspots section Parameter Information Provided Notes Methods To Investigate This table contains five N A suggested methods t
408. rmation the chart and table refresh to show only the instances in that category You can click the back and forward toolbar buttons to move backward and forward through operations you have already performed or you can click the down arrow beside either button and go directly to any of the other operations that you have performed Exploring Memory Analysis Views 141 Exploring the Instances View Baza O Instances 7 022 Heap Other Java Base Libraries java lang Instances 9 963 Heap Other Java Base Libraries java Instances 16 650 Heap Java SE Sample Settings demos memory leakexample Instances 17 104B Use Case Java SE Sample Settings demos memory leakexample You can sort the table by any of the columns The text field above the table allows you to filter the classes that are listed in the table See Filtering Data in the Instances View on page 141 for details Filtering Data in the Instances View When you adjust filters in the Instances view you restrict the number of classes that can be displayed There are two ways to filter the data in the Instances view e You can use the Filter Classes text box above the table to filter by class which narrows what is displayed in the view but does not affect the underlying data set e You can also use the Set Filters dialog to set one or more filters by class Using this method narrows or expands the underlying data set To filter the Instances
409. rom the My Application Filters table Syntax for Filters Filters are case sensitive and blanks are not allowed You can use an asterisk as a wild card it matches any character including the package separator If you specify the code in a shorthand or partial form JProbe automatically changes it to canonical form The change may not be exactly what you want so be sure to verify it JProbe does not check to see whether the filter actually matches anything in your code For example if you make a typing error in your package name nothing in your code will match the filter and so the filter is not applied The following table contains common syntax and describes how JProbe interprets it JProbe User Guide If you type this Canonical form is this Ke And filters are applied to All methods in all classes in all packages including the unnamed package All methods in all classes in the unnamed package only EE All methods in classes named C in any package All methods in the class named C in the unnamed package only M M Method in all classes in all packages P String P String All methods in any class in the package P whose name starts with String and all methods in any class in a package or subpackage of P String matches both P StringArt draw and P Stringent Check English spelling P C do no change Any method whose name starts with
410. rs and a location to save the file The file is saved to the location you indicate Export the chart to CSV Right click in the chart and select Export to CSV In the Export to CSV dialog box specify the export parameters and a location to save the file The file is saved to the location you indicate Show hide table columns Right click and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location 138 JProbe User Guide Task Action Response Sort the table by a Click the column heading for The table is sorted metric the metric To reverse the sort order click the heading again Export the table to Right click in the table and The file contains the table CSV select Export to CSV Typea data separated by a comma file name and click Save It may also contain headers For more information see Changing How JProbe Exports to a CSV File
411. rt The trigger is inserted above the chosen trigger row Removing Triggers You can remove triggers in one of two ways If you think you may use the trigger again you can disable the trigger for the next session Otherwise you can delete the trigger from the trigger table Task Disable a trigger Action Clear the check box beside the trigger s row Response The trigger is ignored during the subsequent analysis Re enable a trigger Select the check box beside the trigger s row The trigger is applied during the subsequent analysis Delete a trigger Right click the button at the left of the trigger s row and choose Delete The row is removed from the table Refining When a Trigger Executes With a basic trigger the trigger executes every time the method is called You can refine this behavior so that a trigger executes based on the number of times a method is called or based on the amount of time that has elapsed 430 JProbe User Guide Example Method based trigger control For example say you want the trigger to become active on the fifth time the method is called and then fire on every other call to the method until the trigger has executed four times To achieve this result you would choose Customize when the trigger action is performed and set the values as n 5 m 2 and x 4 The trigger would fire as follows fire the 5th time the method executes fire th
412. rt table data Right click the table and select Export to CSV See Exporting Table Data toa CSV File on page 442 The file contains the table data separated by a comma It may also contain headers View the execution log from the Runtime Summary view Click Runtime Execution Log 5 The Execution Log dialog box opens and you can review the information by navigating through the available tabs View the history of events since the Console connected to the session Click View Session History amp The Session History dialog box opens In the left pane you can parse the list and review the events that happened since the Console last connected to the session In the right pane you can view additional details for select events Create a report for this view Click Generate Report Er See Generating HTML Reports on page 440 The report contains a summary of the classes and optionally a summary of the methods by class 378 JProbe User Guide Task Action Response Create a support bundle Click Create Support The Support Bundle a Created dialog box Bundle to automatically collect basic support information into a ZIP file for submission to Quest Support appears indicating the location where the ZIP file was saved You can click Copy Bundle Name to copy to the clipboard the name of the bundle that was just created Exploring the Snapshot Bro
413. s To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted Export table data to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by acomma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Create a report for this view Click Export Data to HTML Er The Create Pinned Instances Report dialog box appears For more information see Pinned Instances Report on page 205 160 JProbe User Guide Exploring the Allocations View This section contains the following information e Opening the Allocations View on page 160 e Understanding the Allocations View on page 160 e Interacting with the Allocations View on page 161 Opening the Allocations View The Allocations view can be reached from the Instances view only i
414. s Obsolete listeners also have a greater impact on the performance of your program because as your listener collections grow your program has to propagate events to more and more listener objects 110 JProbe User Guide Lingering Transitional References Long lived objects often contain one or more transitional references that is references that will be reassigned to new objects when the state of the long lived object changes There are two situations to consider In the first case a transitional reference holds onto its object lingers until the method that sets the reference is called again For example consider a print service that has a reference to the current print job The job prints but when it is complete the reference is not freed so the print job remains in memory until the next time you want to print something You can avoid creating these kinds of loitering objects by setting the transitional reference to null when the object is no longer required In the second case a singleton object changes state but one of its transitional references is not updated For example consider an object that maintains information about files in a directory and has references to the largest smallest and most complex files When you change directories for some reason only the references to the largest and smallest files are updated the reference to the most complex file still points to the file in the previous directory This is reall
415. s Report dialog box appears For more information see Instances Report on page 204 Exploring the Dominating Instances View This section contains the following information e Opening the Dominating Instances View on page 148 e Understanding the Dominating Instances View on page 149 e Filtering Data in the Dominating Instances View on page 151 e Interacting with the Dominating Instances View on page 152 Opening the Dominating Instances View The Dominating Instances view can be reached only from the Instances view To open the Dominating Instances view 1 Select a class in the Jnstances view Exploring Memory Analysis Views 149 Exploring the Dominating Instances View 2 Click Dominating Instances ba or Right click the selected row and select the Why Are These Live option from the list The Dominating Instances view appears Understanding the Dominating Instances View The Dominating Instances view allows you to view the instances referred to as dominators that are holding a specific set of instances in the heap and preventing them from being garbage collected Dominators offer a different way of viewing heap data based on reachability analysis If every path from roots in the heap must pass through one instance for example a Widget before getting to another for example a String the former is said to dominate the latter the Widget dominates the String During snapshot proc
416. s g option To obtain line level information your code needs to be compiled in debug mode The following example demonstrates how to modify an Ant compile target to turn on debug mode Assuming that your compile target reads as follows lt target name compile depends init description compile the source gt 256 JProbe User Guide lt Compile code from src and srce test into build lt javac destdir build gt lt src path sre gt lt src path sre test gt lt classpath refid classpath gt lt javac gt lt target gt Create the debug compile target with the following modifications in bold lt target name compile debug depends init description compile the source gt lt Compile code from src and sre test into build lt javac debug 0n destdir build gt lt srce path srce gt lt srce path S sre test gt lt classpath refid classpath gt lt javac gt lt target gt To collect line data 1 Open your configuration file in editing mode 2 Inthe Analysis Type tab click the Filters tab 3 If the method is already listed select Line from the list in the Action column Otherwise add a new row for the method Arrange the rows in the order in which you want the filters applied Caution Filter order is important When two entries refer to the same code the entry lower in the table takes priority 4 To view line data select
417. s methodx which in turn calls methodB most detail You can see the initial calls to encapsulated methods the names of those methods and data associated with each With so much detail to collect however your analysis may take longer and the resulting Call Graph will be more complex Public Method Profiling Data recorded during Performance analysis tracks method invocations included through filters and offers varying levels of detail for method calls that are not included by filters By adjusting the amount of detail on methods that are not included through filters you can adjust the impact JProbe has on your application The number of details you choose to record is directly proportional to the overhead your application may incur The Performance overhead can be greatly reduced by minimizing the number of tracked method invocations Depending on the recording option you enable at runtime or when Learning More About Performance Options 271 Analysis Options for Performance setting up the initial configuration you select between tracking all method invocations included by their filters or only public methods from their filters Public Methods No Encapsulation Triggers for Performance You can select Triggers for your Performance analysis in the Automation tab of the configuration wizard Triggers provide a way to tie JProbe actions to your program s execution You can tell JProbe to perform an action whe
418. s see Creating and Managing Categories Specifying a JProbe Analysis Type You can specify these options during the wizard on the Select a JProbe Analysis screen or afterwards in the Analysis Type tab JProbe offers three analysis types Memory Performance and Coverage You run one analysis at a time The options available for each type of analysis are described in the following sections e Analysis Options for Memory on page 124 e Analysis Options for Performance on page 266 e Analysis Options for Coverage on page 368 Why Three Analysis Types JProbe targets three separate aspects of code correctness and optimization Each of the JProbe three analysis engines are optimized to do a very specific job and to do it very well The different purpose focus and timing of these three types of analysis and the importance of accurate results make it the most sensible technical decision to separate the three engines 46 JProbe User Guide e A Memory analysis requires the highest overhead due to the nature and depth of the object allocation and reference detail being captured When you are doing Memory analysis you should not necessarily care about the overhead because you are not measuring the response time only the memory allocation and reference behavior of your code e Ina Performance analysis accurate execution timing is critical and the Performance tool s low overhead instrumentation does an excellent job of ide
419. s vmid name 13950 demos memory leakexample LeakExample 13886 demos memory network Network 2 Select the application to which you want to attach to and run the command jpattach lt vmid gt where lt vmid gt is the ID number shown at the left of the application For example to attach to the Network demo included with JProbe you need to run jpattach 13886 The command window displays the following output JProbe version 8 3 0 Copyright 2009 Quest Software Inc ALL RIGHTS RESERVED JProbe is a registered trademark of Quest Software Inc Memory analysis attached JProbe Connection Manager is available for Console connections on port 52991 3 On the JProbe Console click Session gt Attach to Session The Attach to Running Session dialog box appears 4 Do one of the following e Ifthe remote host where you ran jpattach and the port appear in the list of found Engines on the JProbe Engines tab while the Product is shown as JMX then select this entry and click OK e Ifthe remote host and port number do not appear in the list of found Engines enter the name of the remote host where you ran jpattach in the Host Name IP Address field and the port number to which jpattach printed in the Port field and click OK Starting Your Application with JProbe Attached 99 Connecting a JProbe Console to an Application Started Without JProbe The JProbe Console connects to the application With the exceptions of the Set
420. s Number of lines in the method that were missed by your test case Total Lines Total number of lines of code in the method 382 JProbe User Guide Column Name Description Missed Conditions Missed Conditions Total Conditions Number of condition outcomes in the method that were missed by your test case Missed Conditions Total Conditions Total number of possible condition outcomes in the method Interacting with the Snapshot Browser You can perform the following tasks Task Examine methods in the application Action Choose the root of the tree Response The lower table displays all methods in the application Examine methods in a package Choose a package branch in the tree The lower table displays all methods in the chosen package Examine methods in a class Choose a class node in the tree The lower table displays all methods in the chosen class Show only the methods with a percentage of missed lines greater than a specified threshold Set the Show Only Methods with Missed fields to Line and to a percentage of missed lines Methods with Missed Lines greater than the chosen percentage are shown The rest of the methods are hidden Show only the methods with a percentage of missed conditions greater than a specified threshold requires condition data Set the Show Only Methods with Missed fields to Conditions and to a
421. s dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted 174 JProbe User Guide Task Export the instance table data to CSV Action Right click in the table and select Export to CSV Type a file name and click Save Response The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Create a report for this view Click Export Data to HTML ery The Create Instance Detail Report dialog box appears For details see Instance Detail Report on page 207 Exploring the Call Traces View When analyzing your application s memory usage it is helpful to view data from several different perspectives If you noticed steep incr
422. s not check to see whether the filter actually matches anything in your code For example if you make a typing error in your package name nothing in your code will match the filter and so the filter is not effective The following table contains common syntax and describes how JProbe interprets it If you type this Canonical formis And filters are applied to this 5 L All methods in all classes in all packages including the unnamed package i 0 All methods in all classes in the unnamed package only C C All methods in classes named C in any package Refining Data Collection and Presentation 435 Creating and Managing Categories If you type this Canonical form is this C aCe M M P String P String P C do no change Pre Pre s y e SEA E foo bar foo bar And filters are applied to All methods in the class named C in the unnamed package only Method in all classes in all packages All methods in any class in the package P whose name starts with String and all methods in any class in a package or subpackage of P String matches both P StringArt draw and P Stringent Check English spelling Any method whose name starts with do in any class named C in any subpackage of P except P directly matches P SubPkg C doGet but does not match P C doNotMatch All methods in all classes starting with Pre in any package as well as any class ina subpackage
423. s that are being called by the root node One of these threads calls the main node Tip When multiple threads execute simultaneously sometimes the elapsed time the program runtime is less than the method execution time The advantage of keeping a separate CCT for each thread becomes clear when using the Prune and Isolate features For more information about Pruning and Isolating data see Isolating Subtrees on page 229 and Pruning Subtrees on page 231 Collected Data It is important to understand the information that is captured in the CCT The following are definitions for time captured in the CCT 224 JProbe User Guide Time Element Definition Method Time The time spent executing code in the method itself it is sometimes called self time for this reason In particular it does not include the time spent executing children it called Current Time The current time measured according to the settings Start Time The current time at the start of the method Cumulative Time Cumulative time for a node in the tree is in general the sum of method time and child time Conceptually it is the total time that the method was active on the stack or the total time between when the method was entered and when it exited Child Time The cumulative time of a node s child nodes The effect of this is that cumulative time accumulates up the tree JProbe notes the start time when a method is entered a
424. s the following sections Analysis Options for COVEraQC essssecssstsseestesteetesneetesnsetesneesteneeateseestesteatesneetesneetenseaneneeanenses 368 368 JProbe User Guide Analysis Options for Coverage You can configure the following settings for your Coverage analysis e Coverage Options e Filters for Coverage e Triggers for Coverage Coverage Options The Analysis Type gt General tab is where you can set the coverage level for a session The coverage level determines the granularity of the data collected during the session Summary Analysis Type Java Application My Code Advanced JProbe Options IVM Analysis Type Memory Analysis Identify and resolve Memory leaks and object cycling Performance Analysis Identify and resolve Bottlenecks and deadlocks Coverage Analysis Identify Unexecuted line of code General Fitters JP automation Coverage Level Method Line Line with Condition The following table describes the Coverage Level options Coverage Description Level Method JProbe tracks calls to the methods in your application respecting any filters that are set Method level data is useful for a first pass analysis especially of large or complex applications You can use the results to identify methods that are missed entirely by your test case Learning More About Coverage Options 369 Analysis Options for Coverage Coverage Level Line Descriptio
425. s the content of the selected JPL file The Settings File Configuration Wizard appears If the tool recognizes the configuration type the Save updated settings to remove errors amp warnings screen appears Continue with step 4 If the tool is unable to recognize the configuration type or if the configuration type is missing from the settings file the Select configuration type or application server screen appears 3 Select the appropriate configuration type from the list and click Next 4 Inthe Save updated settings to remove errors amp warnings screen a Review the list of errors and warnings and attempt to resolve as many issues as possible before running an analysis using the configuration Some common warnings you may encounter include e Missing JPL file for example the JPL file is deleted e Missing value for required field e Filter format error e Trigger format error e Invalid or deprecated JProbe options in a JPL file e Invalid path e Naming conflicts e Other exceptions Note Any deprecated options are ignored when running the configuration file with a newer JProbe version b Review the recommendations presented in the Details section c Select one of the following options for saving the updated settings file e Update values in the selected setting file to update the values in the selected JPL settings file and remove any conversion warnings Don t update values in the selected settings file to
426. s to reflect only the single thread If the method is still a hotspot the selected thread is probably the cause To confirm your assumption you can reset the graph and select the other thread Regardless of the type of application you are analyzing your results would be clearer in the Call Graph if you used the data collection filters Your own methods stand out if you collected method level detail on your own package and left all other code in the encapsulated state Encapsulated data is displayed in the Call Graph as locked nodes that is there is a lock icon on the node or nodes with an asterisk Encapsulated using the Grouped option To identify hotspots along the critical path 1 Inthe Snapshots Navigator double click the Performance snapshot that you want to analyze The Summary view opens providing you with a list of potential hotspots and suggesting the top five focus methods with which you can continue your Performance analysis Note When JDBC data is recorded in the snapshot the Methods tab is displayed on the foreground and the JDBC tab is displayed on the background 2 Inthe Methods tab select the Elapsed Time or the CPU Time option to display the suggested methods based on elapsed time data or on CPU time data provided that the latter is available 3 Select a method in the list right click and select Investigate The Methods view opens 4 Ensure that the Call Graph is set to sort methods by Elapsed Cumulative
427. script backs up opmn xml and copies lt startup_script_name gt _opmn xml to opmn xml After loading the updated opmn xml and starting the OC4J process JProbe restores the opmn xmil file to its original state and deletes the backup Caution Do not end this script using Ctrl C It leaves your application server in an unstable state Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Oracle AS 10g v10 1 3 1 or above When you integrate your configuration JProbe creates a startup script saves your settings file and also makes a copy of the Oracle Application Server s opmn xmil file called lt startup_script_name gt _opmn xml In lt startup_script_name gt _opmn xml java bin data value is replaced with the path to the jplauncher executable Caution You should make a backup copy of the original opmn xm file If you encounter a problem you can manually restore and load your copy of the opmn xm file Usage lt configuration_startup_script gt The startup script backs up opmn xml and copies lt startup_script_name gt _opmn xml to opmn xml After loading the updated opmn xml and starting the OC4J process JProbe restores the o
428. se an initial growth in the heap The memory used by the server is largely out of your control It is worthwhile however to set the size of your original Java heap to be large enough to accommodate the server startup rather than forcing the virtual machine to grow the size of the Java heap dynamically during the startup phase Growth and drop Increasing growth followed by a large drop represents multiple allocations of short lived objects over time that are cleared all at once When the short lived objects fill up the heap the garbage collector frees memory Spikes Steep spikes in the Memory Pools chart represent short lived objects being allocated and garbage collected Frequent garbage collections may cause your program to run more slowly For more information see Tuning Garbage Collection and Finding Over allocations on page 117 GC Data Tab The GC Data tab displays garbage collection data in graphic and table formats The Garbage Collections graph displays a color coded bar for each type of collection and indicates the time that it occurred The total number of GCs per interval is also indicated A table in this view lists garbage collections by collector and indicates for each one the number the length of time it took as well as the cumulative number and time The overhead for all GCs expressed as a percentage of the total time the application ran is indicated above the table Strategies for Memory Analysis 105 Runtim
429. sed time in seconds Tip You can recalculate the data to represent your main thread of execution only by selecting the thread node and isolating its call tree See Focusing on Threads and Methods in the Call Graph and Call Tree on page 249 Time in seconds Thread A 13 f Thread B 6 N method a Your experience of this program is 13 seconds The total cumulative time for the program however is reported as 19 seconds 13 seconds on thread A plus 6 seconds on thread B The difference is the cumulative time contributed by thread B The above example shows that method calls on both threads are accumulated methoda takes one second to execute and is invoked a total of three times so the method time for methodA is 3 seconds 2 seconds from thread A 1 second from thread B The total cumulative time for the program simply adds all the method times together The solution is to isolate the thread of interest in this example thread A in the Call Graph Metrics and Data Collection Filters The following table describes the data collection filters that are available 242 JProbe User Guide Detail Description Method Level Use the Method Level detail filter for your own code The data collected includes metrics for all parts of the method that were executed Encapsulated Use the Encapsulated detail level to hide details on code that you are not interested in analyzing By default data on s
430. seeaseesseeseesneestesneenseraes 437 Communicating ReSults icccscccctcicscnsandiescctccsasecaisesectccsetertactecestotin nttcsateutdirnaeteectecusbectatCsnedsrtnctevcateieasonceaceses 439 Generating HTML Re pont seizes scsecssssererateac wevnaciea nivenivanasamnn nin nanan aKeniten nna AAGA 440 Printing Data to a PDF File s2 00ha aia Ria alee a eee 441 Exporting Table Data toa GSV Filenin 2G ee ea ne sc ee aie 442 Exporting Runtime Ghart Data to a CSM File za nna aaaea e Aaaa E a ATEREA A ie 443 Sanga Chatas am magee a EE 443 Viewing the Execution Log mesnneronragana a T AEA 444 Viewing the Session Historyn seers E E E N E EE AN 445 Generating a Support Bundle rerigen oan E E REN EERS 446 Managing Snap hO S r r ar r rara eaa a arar eaaa a a aaa aar a opnara aeie akian a Setting the Current Snapshot Saving elal EEAO e EE as ae ate E EAEE E A a Loading Saved Snapshots waceosnmsenneetarneansrdiieraimndeia dana ashianaiinnana saminda Unloading Snapshots viv scsracchs Gaal Sle ei ain cede ed Saving Snapshots On EXit esccnnted aie ceisn stink dehredbawi nea a dni aaa Managing Snapshot Transfers Setting PreterenCes a teats tach Sak ada tahecachdsusqcsnusec a r e E ai ed a ai 453 Controlling How JP robe Operates ws stenvttiewcllactlesiiit an iadieagianeie aia asad nites 454 Changing How JProbe Exports to a CSV File eeseeseessneeeeeesessteeiessitesntesstesessesstesaeesnneanieesneesnen 455 Hiding Package Names and Method
431. seestecseessecssessessecseesueessesseeseeeseeseesueeasesueeaseeseeaseestesneesseeseesees Specifying Advanced JProbe Options Advanced JProbe Options aisr iraia aai as ia ap EAA ara aaa aE aa Aa AERE JProbe Pont P c ccitig asi ean aaa a A A A E Snapshot BaSename s 20 20 siesanieeenneaten ana ONENA A A aida 48 Reviewing the Configuration SUMIMALY ssecsecsessessessessessessusssesseesessesaeesessesanssessnssussussusesessteaseaeeaseusenssessateneeetenteateates 49 Saving the CONTIQUIALION 3 nets wens nena iiteacie viii Hine sl daa a 50 Integrating Your Applications iste circa anouuianincaisiuniieinnaunianaiooninaiiad 51 Saving the JProbe Startup Script asccianets cee ie ah ae ah es a ea eel as 51 Executing the JProbe Startup Script sescsesssssssesssssessessessessecsessessessussessessussessuseusseseesesssaseassasesteatsansaneansateaeass 53 Saving the Configuration and Integrating cesessscseseeseseeseesesseesseeseesseeseesesseesucsneeeneeseentesneeneestesneestesneeeeenaesees 55 Running Your Script sssaaa aA So ee EARO EEEE ENET 55 Running Your Script from JProbe Console cs ssessessessessessessessessessessessessuceussassussseseeeacesesaneaesaeeaeeansatsateansaneaceess 55 Running Your Script from JProbe Configuration TOO cessssesesessesseeseesesseeseessesneeseensesnsesseeeestesneeseennesees 57 Running Your Script and Connecting the Console sssssscsessessessesessessessessessessessessesseseessessu
432. ser String String String int false 0 0 removeAliUsers false 0 0 lt clinit gt true 100 0 Per method summary for class demos gamepack framework servlet GlobalServlet Name Method Hit Lines Hit lt init gt true 100 0 init true 100 0 initManagers true 100 0 instantiate_StaticBinding String true 81 8 initGames true 100 0 getServietName false 0 0 setDbMgr Object GamePackManager true 100 0 setUsrMgr Object GamePack Manager true 100 0 setSessMgr Object GamePackManager true 100 0 lt clinit gt true 100 0 Snapshot Browser Report The Snapshot Browser reports include a report header and a summary of the packages in the snapshot You can choose to include progressively more detailed information about the packages classes and methods in the snapshot By default the report includes data on all code in all packages You can choose to exclude some code from the report using report filters 402 JProbe User Guide Create Snapshot Browser Report Output File C 2Probe Snapshats Adventure_Merged J Report Content Application Summary Application and Package Summaries Application Package and Class Summaries All Summaries and Source Data Filter List Pattern j Action KC include Generate File Cancel __ Preview You can create reports that include the following information e Application Summary e Package Summary e Class Summary
433. sesseesssessessececseescsesseeecseeseeesseeasseeseeesseeasseeessesseeasieeaseenaeeaees Opening the Performance Difference VieW cesessessesecsecseesssstecnessteessesseestecneeseeseecseesueesesneeseesneeneentesseens Understanding the Performance Difference View Interacting with the Performance Difference View Setting Preferences for Performance ViCWS scssesessesesesseseesesseeseestesseesuesneeeneesteensesueeseesneesueesuesneesesnesneenseens Reporting Performance Analysis RESUItS s scssessesssersessesersesseserseseeerseseserseeersessteesseeaseesseeaeeneseneesanees 343 Setting ReportOptions sc wee stain cea a an ee ee ee A R 344 Runtime Summary Report ccscissisienimrauiniisanshsanhuais tact Guuaniahuirsdaidbansinaladeisindie 344 SUMMAry REPOlbe tata araetviwsacateeichned cate cinta es even dadaeinch ET 345 Methods Report siisciacianudnaiannipaianaisinesooseaaianuununianaiannainannnunianainnianniad 346 Method Detail Report fs0 c2 aaa Sa ee ee eee 347 Performance Difference Report i csitinsyaneittndiecin iuadeniiiay a 347 Troubleshooting Performance Analysis Results s ssssssssessssseeeersseesersseeeeersseeeenrsneetsnrsneeeeasennseennaneeees 349 Performance Troubleshooting Tips ssiiwicsteaseciissiniciisin nadia canalsuriiahihiniianaiiunianniae 350 Performance Session RUNtIMG TIPS scsi scaiautead deni cain cine A R 351 Performance Help MOSSAGES gs cscicssesscseisssnsassisasiscbuaseassaacesssoassin
434. settings file and the jython script are saved to the same directory do not separate these files When you run the JProbe generated startup script you can specify parameters that you usually pass to the WebSphere startup script Do not however specify the server name on the command line because it is set in the configuration If you do specify a server name it is ignored To choose a different server copy the current configuration change the server name and integrate Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly Usage lt configuration_startup_script gt options where options are clean Restores WebSphere to a pre JProbe state after an abnormal termination e nowait e quiet e logfile lt filename gt e replacelog e trace e script lt script filename gt background e timeout lt seconds gt e statusport lt portnumber gt e recovery e help Before starting the server the startup script updates the server configuration settings to use jpjava executable instead of java executable After starting the server the script restores the server configuration settings 84 JProbe User Guide If
435. seussesseseissateaseateaneaseateanes 58 Managing Gontigurations scisasih ahaha ute Bila ile Ri tae ieee a eel R 59 Pre Defined Demo Configurations eesscsessecssssessscnsssecssecsessesseesseessesseeseestesneensecnsesneeatecnteatecneeateeanesseensesiss 59 Copying a Configurations nsee a EE E TET A N anaes 59 Editing a Configuration si r ccnacsnsanaeting anieaion E ARTT 60 Deleting a Configura N sccs sccctetsaciancsteedsaccasenuitesstaaceaepenceassepacstessassatesaaatssarearieanscaneteneasatea toda ancients atials 61 Creating a Configuration from a Settings File ceseseseesesssesseesessesneeseeneeeneesneeseeneeseesteseesssueessesneesieesesees 61 Renaming a Settings File aaae erara aere A Aaaa Ea eE TE NA E T EE 63 What Can I Do if My Application Server Is Not Supported oo eeeseeseesteestesstestecstesseesseesteeseesneeseesneesusentesnsentesnteeneentensees 64 Creating a Configuration for an Unsupported Application Server c cssscsssessesesstessessesesessesseesteaneseesnsetesneanes 65 Requesting a Module from Quest SUpport sssssessscssssesssssssssssessessessussessessssessussesssseaeesssaneaeeaneateasesteaneaneaneaeens 67 Modifying an Application Server Startup Script ccccscsecsscssssessecsessessessessessessessssusseseseesssaseassaneanesesseseteateateateass 68 Starting Your Application with JProbe Attached Running a Java Application with JProbe 4 Running a Java EE Application with JProbe ses
436. sis fewer object allocations and releases are recorded during the data recording which results in fewer instances to analyze Inthe case of a performance analysis less method entry exits are tracked which results in a smaller call tree graph e Inthe case of a coverage analysis less method entry exits are tracked which results in fewer classes packages that are displayed To add application filters 1 2 Type a category name in the Category Program name text box In the My Application Filters table click an empty cell and do one of the following e Click the cell Browse button tJ navigate to a custom package and click Open e Double click the cell and enter the package name The package is displayed in the cell and the Action is set to Include If the check box at the beginning of the row is selected this indicates that the filter is enabled Repeat for other custom packages To exclude some code in an included package click in an empty cell below the package row and do one of the following e To select a subpackage or class click the cell browse button navigate to the package or class and click Open Set the Action to Exclude e To select a method click the cell browse button and navigate to the class containing the method When you select the class you can expand its list of methods Locate and double click the desired method Set the Action to Exclude e Enter the package class or method no signa
437. sis 263 Investigating Your Results while done synchronized queue_ if queue_ isEmpty queue_ wait item queue_ remove useItem item Suppose produceItem throws an exception while the consumer thread is executing queue_ wait and the producer thread terminates The consumer thread is never notified and the program hangs It is possible however that the program could continue in this situation For example the main thread might notice the terminated producer thread and create a new producer in which case the program would continue A hang can also be caused by problematic thread priority For example the thread priority might keep a lower priority thread from ever running possibly while the lower priority thread holds a resource required by a higher priority thread called a priority inversion Performance does not detect these hangs as deadlocks Finding Deadlocks JProbe can detect deadlocks on demand To detect deadlocks while running a Performance analysis 1 In JProbe connect to a running session of your program configured to execute a Performance analysis 2 Inthe Runtime Summary view click the Threads tab 3 Click Detect Deadlock If deadlocks are detected the list of threads in the lower pane is updated to show only the threads involved in the deadlock If no deadlock is detected a message is displayed to that effect 264 JProbe User Guide welcome F R
438. sisiini NEE E eeni ata EEEE TEE ATENAS i Refining Data Collection and Presentation sssssssessssuressuneuuunesnunuuuunnonunununnnonunuunnnnonununnnnnonnnnnnnnnonnnnnnnnnenann 421 Refining Data Collection Filters Tiris inara a E A TA cheese 422 Adding FIRSTS iksiri iri apriete rarae r laane E Naaa ANARA EEan A RE EENAA EAEE EE EPERERA EENAA AA 422 Changing Filter Exec tion Qrdersnssn iaar r E R a 424 Removing Fitos measann aa R E neers 424 Setting Triggers Table of Contents 13 Adding Triggers iaaa a a ARR Setting Advanced Trigger Controls Changing Trigger Execution Order Removing Triggers as a aaa E AEE a a aa aiai Refining When a Trigger Executes niiina n ai a aaa Adding a Trigger in the Body of a Method cessscsessssssssessessessessessessseesessseseesseseeseeseessenseneesteaseateseenseatesentesness 431 Creating and Managing Categories cecsesesssesseseesesseesseesesseesneestesneeseensesnsesseniesneesnesseesnsesesuesneennseneeseenieeneennee 432 Adding Categories sesicrecccs directorial ad n a a a aaa 432 Adding Folders arretea anakara aeaaaee a EATE Naa E at aniseed unin aianninionnianiinalianta 436 Editing Cat goriesten ana A Ra E 436 Renaming Categories or Folders scsssessscessecssesnsestecseesseesecstesneesecseesneeatecnteaeeaeentesneeasesseestesteeseeatesneenseeaes 437 Deleting Categories or FOIMCrs seesseecsessessesssecsessecstecnesstecseesseeseesseessesseesseeseeeasecntesseesees
439. solve Bottlenecks and deadlocks Coverage Analysis Identify Unexecuted line of code P General A Fiters HP Automation v Data Collection Filters package class method Action Add My Code V demos coverage adventure include lt Filters are defined in the same way for all the analysis tools however the filter actions differ among the analysis tools This section summarizes the filter actions for a Coverage analysis To learn how to create filters see Refining Data Collection Filters on page 422 Filter Action Include Description Default By default all classes and methods in the application except for native and abstract methods are included in the analysis and do not require explicit filters You can use this action to include methods from a class that you have excluded It also gives you the flexibility to restore a previously excluded class or method to the analysis without deleting the filter Exclude You can tell JProbe to ignore code in a specified package class or method The resulting snapshot contains no data on excluded classes and methods Native By default native JNI methods are excluded from the Coverage analysis Select this action for specific native methods that you want to include Note Avoid assigning this action to non native methods or to packages or classes that contain non native methods If you do the resulting data ma
440. ssssesssesesssseseeseenesseesessteseeaeesesseenesassneeseseenesnseaesseeetsaneaeenesineeeeaneeees 101 Recording Datta st scsecen ise AnieniooneAwienar 2 Me Avor a dead A A ait Aedeagal 102 Runtime Summary 3 2 4c ell eevee andes eee iar Meda T 102 Memory Pools Tab is ifccist avuiaw aiaativaa iiesiieneit A T O nin aetiion 103 COAPTE Fl o ETA Oo pI RE EP OPE TRAE rt 104 6 JProbe User Guide Heap Data Tabiano E eran aaa oA Chien 105 Running a Memory AnalySis scscsssscscsssecsessesesersseerseeeserseeseeseserseeessersesseserseeseeeeseeaeeesseeaeeesseeaeeessneateessneas 107 Finding Memory Leaks asccssiscieveasasctsasaisssaetisasssesifiinasaseosaniaauuianuapiuiiannaienaucianaiienisaiiaiay 108 Understanding Loitering Objects seesseseesessecssecneestecsesssesseecneeasecseesseessesseesseesesneesueeaseenseateenseateenseaneeateenees 108 Setting the Recording LeVert eap anean sapaa ARAA EEEE mann dendaikn aa iaia 110 Performing a Basic Memory Analysis ineen kee a a a A E E 112 Tuning Garbage Collection and Finding Over allocations ccseessesesesseseeseesesseeseesseneeseeseesieseentesesniesneeneneee 117 Understanding Garbage Collection ISSUCS esscseessesssessesseeseesneessesseeseeseesneessesseeseeseentesnseateenseateeneenneentennees 117 Reducing Over allocation Of Objects ss csessscsecsecsessessessessessessneseesnsenssessnseussaneseesnsaneansstsensenesnssteaneensaneeneeneenes 119 Examining Garbage Collecti
441. ssssssssssssssssesssesessssssesssssessseseeseeseenesseenessteeteaneetnassenesaeeseentans 265 Analysis Options for Perlormances c i ccsscsevcessissccsateectessitecrsesteineetidennsusteviestvaedecsit a iie Timing Settings for Performante ite adesulh esate edith iain ie eel etl eae f JDBC Component Instrumentation Filters for Perlonmance isis iann aaa AN AE NEA A ER Automation Settings for Performance sscsccsscsessecsessessessessessssessussussessussussuseueesesasesesaseaseateaseaneassaneaneateaneaseesees 268 Triggers for Pertormance isin ew dia ai E E acai weariad 271 Exploring Performance Analysis Views Exploring the Runtime Summary View Understanding the Runtime Summary View csceseesseseseesecseesnsesneestesneestecneeseeseeentesneestesseesneenesatsnesneensenses 274 Interacting with the Performance Runtime Summary VieW s csecsecssssecsessessessssesseesuesstsseesesstsaesatssnsansenseteateess 279 Exploring the Summary VEW a er eraat aparinata apaan Epa aaa A Aaaa EATARRA Sana fakata ESA SAARE a EAA ARARA 283 Opening the S mmary ViCW ccseesssecesseecsesssessescnsssnesstecuessesstecseestecsteseestessessseestesessecntesseestesaeeateeateaeersteenees 283 Understanding the Summary ViW cccscssssssssessssessesssssessessessessessessesseseusesssuseusesseaceuesansateaneaneaneatsaneaneateeneaeess 285 10 JProbe User Guide Interacting with the S mmani Vie Ws cust acdie anaes aa ea E dha
442. st Action Select a class from the table and click Dominating Instances iad Select a class from the table right click and select the Why Are These Live option from Response The Dominating Instances view opens showing what instances are dominating pinning the selected instance set For more information see Exploring the Dominating Instances View on page 148 Exploring Memory Analysis Views 145 Exploring the Instances View Task See the instances that are pinned by the selection Action Select a class from the table and click Pinned Instances be OR Select a class from the table right click and select the What Do These Pin option from the list Response The Pinned Instances view opens showing the instances dominated pinned by the selected instance set For more information see Exploring the Pinned Instances View on page 155 See the instances for a class or classes Select a class from the table and click Instance Detail View f Note This action is available only if the selected classes have instances that are live in the heap The Instance Detail view opens with all the instances for the selected class listed in the upper pane For more information see Exploring the Instance Detail View on page 166 View the merged allocation points for the instances of one or more classes Set Investigate By to Recorded Count Memory or
443. st case did not follow a particular conditional branch In this case you should run the test cases again while tracking condition coverage Determining Overall Coverage First you create a baseline snapshot that lists all the classes in your application Then you run a Coverage analysis on each test case being careful to use the same class files coverage level method or line and filter settings as the baseline snapshot When you merge the test case snapshots with the baseline snapshot the merged snapshot tells you the percentage of overall coverage that your test suite provides You can generate reports to communicate your results to management and the rest of your team Modifying a Test Suite As you analyze your test suite you may find areas of code that were not exercised Sometimes it may not be clear if you need to test that code or not To find out you can generate reports and show them to the software developer responsible for the code The software developer should be able to tell you if the code needs to be tested or if it is obsolete Creating a New Test Suite If you are starting your test suite from scratch a good first strategy is to talk to the software development team and determine which use cases are most attempted by your users Create test cases for those use cases first and then use JProbe to ensure your test cases cover all the code affected by those use cases This strategy ensures that at a minimum your test suit
444. stomizing the Heap Graph sivsciieis cence daa cea ainda aaa etd Interacting with the Heap Graph er 2ianesanen are cavenraiani inanin e ER ERRERA Exploring the Leak Doctor View Opening theLeak Doctors osean sinnini aiid E n EN Understanding the Leak DOTr sai a aaae quinine a E AEA A ENARE Interacting with the Leak Doctor ViOW sssssecsessessessessessessessesseeseesesesesuecseesesaeeneesesseeseeseeseeneeseeateaeeseeaseateaeneeess Exploring the Memory Source VOW a 3 sssccss eas ceatetueeastenectasdasistnsasestusarsateseetauaehaenatesmean iida tip saipa anidrita eaaa 8 JProbe User Guide Opening the Source VieW fai3 anaieiemtinediies anaviintaw dearitilaalanadnwedinn 197 Interacting with the Source ViCW ceseessessessesseesseesecstecsesstecseesneestecseesseeseesseesteeseesieesueeateenssnteeneeaeeneeaneeateenees 197 Exploring the Memory Difference ViGW c csccscsessssesecessesesseeseseeesseeeeseeseeesseeesseeseesaseesseeateassnsesseeaeeesseeasensaneesaeeaes 197 Opening the Memory Difference View escsesseseeseseesesseeseesesseesnsestesneesseestesneestesneesessnesneesneeeesnsensesneenaeees 198 Understanding the Memory Difference View csscsssssssssteseesteeteseeenssesseeneeassneeatsateatenteatsanesnsensseesneaneaneneennes 199 Using Filters in the Memory Difference View ssssssssessteseesteetssteesstesseestsnsestsateateateateanentesteatenneaneanentennes 200 Interacting with the Memory Differenc
445. subpackage that starts with s matches both java rmi server RMISocketFactory com quest say and java security cert X509CRLEntry foo bar foo bar All methods in the class bar in package foo This is likely not what you wanted See the next example foo bar foo bar All methods in all classes in foo bar or any of its subpackages foo bar no change Methods whose names start with t and end String t e in ein foo bar String matches both toUpperCase and toLowerCase 408 JProbe User Guide 20 Troubleshooting Coverage Analysis Results As you work with JProbe you may need help to get the most out of your analysis sessions This chapter presents a few tips for solving some Coverage analysis problems and for running sessions efficiently For Coverage analysis issues that are not covered here refer to Quest s technical support This chapter contains the following sections Coverage ToNeshooting TIPS senini danna eaea ieai 410 410 JProbe User Guide Coverage Troubleshooting Tips Lines of code with identical statements show different results Identical source statements may be compiled differently by javac or the line table generated by javac may have assigned the bytecode to source lines differently For more information see Nonintuitive Results for Some Source Code on page 393 A statement that think should be executable is non executable or vice versa Your javac comp
446. t JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Usage lt configuration_startup_script gt commands where commands are e run Start Catalina with JProbe in the current window e run security Start in the current window with security manager start Start Catalina with JProbe in a separate window e start security Start in a separate window with security manager Starting Your Application with JProbe Attached 81 Runtime Notes for Application Servers BEA Systems WebLogic v9 x 8 x You run the JProbe generated startup script in the same way that you run your WebLogic startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables and the same command line arguments Usage lt configuration_startup_script gt BEA Systems WebLogic Portal v9 x You run the JProbe generated startup script in the same way that you run your WebLogic startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process o
447. t if the resized columns do not fill the table the width of all columns is increased to fill the empty space Setting Preferences for Coverage Views You set preferences in the Options dialog box For more information see Setting Preferences on page 453 This section contains the following information e Displaying Percentages in Bar Charts e Filtering Out Catch Block Data e Customizing the Colors Used to Display Results Exploring Coverage Analysis Views 391 Setting Preferences for Coverage Views Displaying Percentages in Bar Charts By default JProbe displays percentage based metrics in the context of a bar chart You can hide the bar charts To hide bar charts 1 Click the iy Options toolbar button or choose Tools gt Options 2 Choose Data Display gt Coverage 3 To hide the bar charts clear the Show Bar Chart for Percentage Metrics check box 4 Click OK When you return to a Coverage view the bar charts are hidden Filtering Out Catch Block Data The data for catch blocks is included in the following metrics e Missed Lines e Missed Lines e Total Lines e Missed Conditions e Missed Conditions e Total Conditions Catch blocks however are often impossible to test If you choose not to test them you may find it easier to interpret your results without the catch block data inflating the missed values You can set a filter to remove the data from the calculations To filter out catch bloc
448. t Tune SQL Statement The SQL Optimizer for Oracle opens with the SQL text already inserted Note You can repeat this step for all SQL statements in the list with the exception of batch queries 4 Run the SQL Optimizer for Oracle to tune your SQL statement and identify the optimal one You can update your SQL statement s in the original source by using the tuned statements then run your application again with JProbe The What Next section guides you to choose a connection or statement to investigate The graph node associated with the connection or statement is the focus node for the Methods view that opens when you continue your Performance analysis Interacting with the Summary View Several operations are available from the Summary view s tabs from the toolbar buttons or the toolbar menu For details see the following topics e Exploring the Summary View Toolbar e Exploring the Methods Tab e Exploring the JDBC Tab Exploring the Summary View Toolbar The following table lists the operations that you can perform from the Summary view toolbar Task Action Response View the execution Click Execution Log amp The Execution Log log from the dialog box opens and you Summary view can review the information by navigating through the available tabs Exploring Performance Analysis Views 293 Exploring the Summary View Task Create a report for this view Action Click Generate Report
449. t in the method H1 and H4 is attributed to V2 s method time and the time for H2 H5 and H3 is attributed to V3 s method time 254 JProbe User Guide To encapsulate data on methods in your application 1 In the JProbe Console select a configuration from the Configurations Navigator The lt Configuration_File_Name jpl gt tab appears in the View area Click Edit The lt Configuration_File_Name jpl gt tab is placed in editing mode On the Analysis Type tab select the Performance Analysis check box On the Automation tab move the slider to select one of the options available in the Initial Recording at JVM Start section e Data Recording Off e Public Methods No Encapsulation e No Encapsulation e Grouped Encapsulation e Full Encapsulation Note Unless you need to track things from the start of the JVM it is recommended to set the initial recording to Data Recording Off and use triggers to start recording at a relevant time Click Save Start a Performance analysis session To modify the recording level during the analysis session click Set Recording Level on the Runtime Summary view s toolbar The Set Recording Level dialog box appears Select one of the encapsulation options and click OK Running a Performance Analysis 255 Investigating Your Results Method level Dat ee i Pn a at gt ore E E a ee we jj O M panpa E lated on Collecting Data on Lines of Code
450. t is referring to the instance that you are investigating The type of the reference The amount of memory that would be released if the object is removed from the heap The field in the referring instance that refers to the referenced instance The References tree provides the following information Field Name Type Keep Alive Size The field in the referring instance that refers to the referenced instance The type of the reference The amount of memory that would be released if the object is removed from the heap Name Exploring Memory Analysis Views 169 Exploring the Instance Detail View The name of the field that references the instance in the tree The name of each field in the Referrers References trees is prefixed by an icon which provides information about that field as described in the following table Icon Name Type Recorded Instance created when the recording was enabled On Not Recorded Instance created when the recording was disabled Off Root Root instance Static Root Static root instance ES Cycle Instance that refers to itself circular reference Note The status bar also displays the most common icons and short descriptions of what they represent Dominators The Dominators tab contains an expanded tree table that contains a chain of referrers back to a root highlighting dominating instances ba The table provides the follow
451. t time which could be problematic User Code A transition from a user class to another component which might be worth investigation A number of instances in the heap will be pinned by roots the different types are described below e JNI Global Roots happen when native code creates a global reference to an instance e JNI Local Roots happen when native code is called with instances in the heap e Stack Local Roots happen when a method is invoked they pin the local variables of the method e Sticky System Class Roots happen for a number of classes in the system that the VM does not want to go away e Monitor Roots happen when something acquires a lock e Thread Roots exist for each thread running in the system e Other not in any of the above categories 194 JProbe User Guide Direction of analysis lt lt Possible cause What do we call this Loitering object Old Objects New Objects When you choose a reference in the upper pane and click the ts Remove Edge toolbar button the reference is removed from the heap and added to the table to the lower pane If removing this reference frees the loitering object the following message appears Instance Eligible for Garbage Collection Ed e Congratulations the instance can be garbage collected Your next step is to free the removed edges in your application code You now know t
452. t to start your application server The startup script field is pre populated with the latest startup script saved for the selected configuration Note Ifthe startup script file needs any environment settings you should set these environment settings before executing the jpconfig or the jpconsole application For example if the script file needs a JAVA_HOME value you should first set this value in your environment then run jpconfig or jpconsole and finally proceed to run the startup script file e Optional Type in any script parameters that should be used when launching the script e Select your working directory where the startup script resides e Optional Select the Close syscon window on exit check box to automatically close the jpsyscon window after the termination of the launched script if the 58 JProbe User Guide program exits normally If this check box is cleared or if the program exits with a non zero status you must press ENTER to close the jpsyscon window Note This option is global and is not saved against each configuration value e Optional Select the Close Create Edit Settings tool while executing JProbe startup script check box to automatically close the Configuration tool window after the termination of the launched script if the program exits normally 4 Click Run The selected application server launches You can now open a JProbe Console connect it to the running application and start analyzing your a
453. ta Right click the table and choose Export to CSV See Exporting Table Data to a CSV File on page 442 The file contains the table data separated by a comma It may also contain headers 384 JProbe User Guide Task Action Response Create a report for this view The report contains the level of detail chosen in the Create Coverage Report dialog box Click Sal Export Data to HTML See Generating HTML Reports on page 440 Exploring the Coverage Source View The Source view displays the source code for a chosen method This section contains the following information e Opening the Source Code e Understanding the Data in a Source View e Interacting with the Source View Note You need access to the application s source code to use this view Opening the Source Code You can have multiple Source views open Each view is labeled with the name of the source code file displayed within it To view source code 1 From the Snapshot Browser do one of the following e Double click a package class or method e Select a package class or method and click the a View Source toolbar button 2 If prompted locate the source code file for the class The source code is displayed in a Source view Exploring Coverage Analysis Views Exploring the Coverage Source View Understanding the Data in a Source View In a Source view your code is color coded to highlight missed
454. ta in the Instances View on page 141 The classes in the table reflect the filtering expression s you have entered Calculate Actual Keep Alive Size The sign in the Keep Alive Size column indicates an estimated value for this metric To calculate the actual size select a row right click and select Calculate Actual Keep Alive Size from the list The Keep Alive Size column is updated with the actual Keep Alive Size value Set your primary investigative path Select one of the following options from the Investigate By list e To investigate recorded objects select Recorded Count Memory e To investigate the whole heap select Heap Count Memory e To investigate garbage data select Dead Count Memory The horizontal bar graph is re drawn to reflect your choice Exploring Memory Analysis Views 147 Exploring the Instances View Task Save the chart as a PNG or JPEG file Action Right click in the chart and select Save Chart As then select the graphic file type PNG or JPEG in which you want to save the chart In the Export Image dialog box specify the export parameters and a location to save the file Response The file is saved to the location you indicate Export the chart to CSV Right click in the chart and select Export to CSV In the Export to CSV dialog box specify the export parameters and a location to save the file The file is saved to the locat
455. tabs One displays the referrers and references for the selected instance the second displays the dominators of the selected instance and the third displays the allocation stack trace for the selected instance welcome amp Runtime Summary B Instances B Instance Detail AREE Name Size Allocation Time Referrers 4 References Keep Alive Size Allocated at java lang String Ox2d4F 1 2 a java lang String 0x1b82 24 1 2 96 Trace was nat collected java lang String 0x2ae1 24 1 2 24 Trace was not collected java lang String Ox2d49 24 1 2 48 Trace was not collected java lang String Ox2d4e 24 1 2 72 Trace was not collected java lang String Oxe8f 24 it 2 24 Trace was not collected java lang String 0x1b81 24 1 2 96 Trace was not collected java lang String 0x2ae0 24 1 2 24 Trace was not collected java lang String Ox2baF 24 1 2 72 Trace was not collected java lang String 0x259c 24 1 2 96 Trace was not collected java lang String Oxdbf 24 1 2 24 Trace was not collected Referrers References Dominators Trace Referrers References Name Type Keep Alive Size Field Name Field Name Type Keep Alive Size Name E amp lt class gt sun text re CONSTANT_POOL 5 008 lt constant pool 36 gt ic amp lt class java lang St CLASS 456 lt class gt java lang String m value FIELD 64 char 0x2dc6 Instance List The upper pane contains t
456. tatements to display all the statements represented by that row Method Displays the name of the JDBC method used to execute the query Calls Displays how many times the JDBC method was called CPU Cml Displays the CPU cumulative time Elap Cml Displays the elapsed cumulative time By double clicking a row in the SQL Statements section you open the Methods view with the graph associated with the row as the focus node By selecting a row in the SQL Statements section right clicking and selecting the appropriate option you can investigate where the selected statements were executed or the statement s SQL was collected The Show only DML check box is by default selected restricting the SQL statements shown to the appropriate statement types that is only a subset of SQL statements known as DML Data Manipulation Language is displayed To display all collected SQL statements clear the Show only DML check box DML statement types are appropriate for tuning using Quest s SQL Optimizer for Oracle To tune an SQL statement 1 Install Quest s SQL Optimizer for Oracle on the machine that will run the JProbe Console Note For information about this product refer to the SQL Optimizer for Oracle documentation set 292 JProbe User Guide 2 Open the performance snapshot containing SQL statements in the JProbe Console 3 In the Summary view gt JDBC tab right click the statement that needs tuning and selec
457. te or URL e n netboot lt url gt Boot from net with the given URL as base JBoss v5 x You run the JProbe generated startup script in the same way that you run your JBoss startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables such as JAVA_HOME and JAVA_OPTS and the same command line arguments Usage lt configuration_startup_script gt options where options are e c configuration lt name gt Set the server configuration name e p patchdir lt dir gt Set the patch directory must be absolute or URL e n netboot lt url gt Boot from net with the given URL as base Open Source GlassFish v2 UR2 When you integrate your configuration JProbe creates a startup script resaves your settings file and also makes a copy of the GlassFish domain xml file called lt configuration_startup_script_name gt _domain xml In lt configuration_startup_script_name gt _domain xml calls to java are replaced with calls to jplauncher Caution You should make a backup copy of the original domain xm file If you encounter a problem you can manually restore and loa
458. tennee 135 Exploring the Instances VieW Ase whieh et cas AU UN ed el ee 138 Opening the Instances View sesssesssssessesssssessessessessessessesseseucsnsenssussussussussuseuesessasesesseeateateatsateaeseneaneateatsateenees 138 Understanding the Instances View csessssessessessesseesessesetestscsestessesseesessessesseeseenesseeaseneeseeneeseeneenteseeaeesesnenees 139 Filtering Data inthe Instances VIOW cscsccccanie resins nine ratncanencnowanirainacienn inna 141 Interacting with the Instances ViW c scccccsesessecsesecseseesesseseseesseseesesaseeeneeessesseeneseeeseeseneaseeseeeesaseesineaseneateeseneees 144 Exploring the Dominating Instances View s ssssscssssecessesssesesseseesseeeesetesnesesssseneeneeseeneeneeseeaeesseeseaeeateateateaeeseeatesaess 148 Opening the Dominating Instances View escsseesssesecsecsessesseecnesstecseesuesstecsessseeseesseesesseesueeseesneestesneeseeneesneess 148 Understanding the Dominating Instances View esssssseseseeseeseesessnseesseseesesseenteetsneententeatenteeteaneetenneanentennes 149 Filtering Data in the Dominating Instances View c sescccsesecsesseseseseeseseseseeseeeseesssesatensseeseeeseneeseeeessneateesateaees 151 Interacting with the Dominating Instances View sssssssseseestestestestesesneenesstsstestsateatentestsanesteeseneseenneaneneennes 152 Exploring the Pinned Instances Vio Wasossarmisnenamasirmnieenaninionginanai maiaa 155 Table of Contents
459. ter the list to show only your methods first and remove encapsulated third party methods 244 JProbe User Guide from the view You can then sort the list in terms of each of the metrics Your most expensive methods appear at the top of the list or the bottom if sorted in ascending order Compare the results against your expectations If you are analyzing a standalone application see Finding Hotspots by Navigating the Call Graph on page 247 To find the most expensive methods using the Call Graph method list 1 Ifthe snapshot is not already open in the Snapshots Navigator double click the performance snapshot you want to analyze The Summary view opens 2 Double click the method of interest The Methods view opens with the selected method as the focus method 3 Enter your package name in the Filter Methods field The list displays only the methods from your package 4 If the list is not sorted by Method Time click the Method Time column heading twice Your most expensive methods in terms of method time appear at the top 5 Assess each of the top ten methods looking for ones that take more time than you expected 6 After you identify one or more hotspots go to Investigating Performance Bottlenecks on page 248 Finding Hotspots in the Critical Path The simple method described above has two drawbacks One is that the hotspots may only be hot on some call paths This is not apparent from the method list a
460. ter the session ends you can open the snapshots in the JProbe Console Alternatively you can attach a Console take a snapshots and when finished detach and close the Console Do not collect stack traces or garbage data if you do not need them Collecting stack traces can be expensive and may not be needed for an initial investigation If you do need traces or garbage data use the Set Recording Level dialog to collect traces for only the classes you need Troubleshooting Memory Analysis Results 215 Memory Help Messages Reduce load on your program before taking snapshots Taking a snapshot is an expensive operation in the JVM It will be faster and more likely to succeed if your program is idle Memory Help Messages When you try to execute an action that is not enabled the JProbe Console usually displays a message explaining why that action is currently disabled and suggests what you need to do to resolve the situation This section provides a list of memory help message you may encounter while doing a Memory Analysis using JProbe e No Heap Data on page 215 e No Stack Trace Data on page 216 e Operation Not Valid for Categories on page 216 No Heap Data You can execute this operation only if your snapshot contains heap data A snapshot with no heap data happens very rarely It may occur in one of the following cases e You have imported into JProbe a snapshot taken with no heap information No workaround e
461. terprise 472 node locked 464 per seat 468 Line Level filter 241 line level data g compiler 415 improving 415 lines of code collecting data on 255 lingering transitional references 110 listeners obsolete 109 loading snapshots 449 loitering objects 108 M main class specifying for a Java application 32 managing configurations 59 licenses 463 measuring performance improvements 258 Memory Allocations Detail view 163 Allocations view 160 analysis options 124 Index 479 analysis strategies 101 analysis performing a basic 112 automation 126 Call Traces view 174 customizing the Call Traces view 177 187 Difference view 197 199 Dominating Instances view 148 filters 124 GC Data 104 134 Heap Data 105 134 heap dump importing 127 Heap Graph view 185 help messages 215 importing IBM Dumps 127 129 Instance Detail view 166 Instances view 138 Leak Doctor view 190 Memory Pools 103 133 Merged Allocation Points view 181 Pinned Instances view 155 recording data 102 reports 203 running an analysis 107 Runtime Summary view 102 132 session runtime tips 214 Set Recording Level 110 Source view 196 triggers 125 troubleshooting tips 212 views setting preferences for 202 Memory Difference view 197 interacting with 200 opening the 198 percentages 200 using filters with 200 values 199 memory leaks finding 108 Memory Pools 103 chart 221 configuring 456 interpreting the graph 103 480 JProbe User Guide Memory
462. the 426 JProbe User Guide action takes place that is on entry to or exit from the method The action corresponds to an action that you would normally perform from the JProbe Console Caution Triggers are global which means that they are not local to methods or even threads Each trigger occurs independently when the method is encountered and affects the whole program The procedure for adding triggers is the same for all the analysis tools however the trigger actions are a little different among the tools Note You can add triggers during the wizard on the Specify JProbe Analysis screen or when editing the configuration on the Analysis Type tab To add a trigger 1 On the screen tab where you define the analysis type click the Automation tab and click Add Trigger The New Trigger Definition dialog box appears The upper part of this dialog box is the same for all types of analysis but the lower part contains different options depending on which type of analysis you are performing and or the action that you choose 2 Do one of the following e Select a trigger method from classes defined in your workspace projects a Click the browse button beside the Trigger On Method text box The Select Trigger Method dialog box appears Select Trigger Method Select a trigger method From classes in workspace projects ted Bees E LL_MyFistJavaProject E TestJavaProject b Navigate to and select a class
463. the CCT this means that a node for that method has a parent or ancestor node for the same method JProbe saves space by using special recursive nodes For example consider the following CCT for Factorial Strategies for Performance Analysis 233 Understanding Performance Data main 1 S Factorial 1 _ Factorial 1 To simplify this tree multiple nodes are replaced with a single recursive node S Factorial 1 Factorial 1 Recursive example showing call count A more complex example like the classic Towers of Hanoi puzzle with 4 disks the call tree showing call count might look like the following example MoveDisks 1 MoveDisks 1 rl MoveDisks 2 Printin 1 MoveDisk 2 FindEmpty 2 MoveDisks 4 Printin 2 MoveDisk 4 FindEmpty 4 MoveDisks 8 Printtn 4 MoveDisk 8 MoveDisk 1 Ht Printin 8 Towers of Hanio example Cumulative Time and Recursion With our original definition of cumulative time self time and the cumulative time of children the cumulative time for MoveDisks is encountered twice If the Recursion example displayed method time instead of call count the cumulative time for MoveDisk would be 30 15 15 Therefore the cumulative time for MoveDisk is 30 74 14 15 66 234 JProbe User Guide Essentially the 14 from MoveDisks R is double counted because it is included in both the method time an
464. the Configuration Type E3 Apache BEA Systems BEA WebLogic v9 x BEA WebLogic Portal v9 x PSEA WebLogic v10 0 n IBM E JBoss 8 Java Application Open Source f Oracle EA Settings File SpringSource Required Field Behind the scenes a configuration is stored as two files a settings file and a startup script For some application servers JProbe may also copy edit and later restore a configuration file for the application server For more information see Integrating Your Application on page 51 After you successfully run your first analysis you may want to run other types of analyses or use different options Rather than changing your configuration you could create a new configuration based on the existing configuration For more information see Copying a Configuration on page 59 Configuring JProbe to Analyze Your Application 31 Specifying Your Java EE Application Server Specifying Your Java EE Application Server You can specify these options during the wizard on the lt Java EE Application Server Name gt screen or afterwards in the lt Java EE Application Server Name gt tab JProbe needs to know about your installation of the application server If your application server has a startup script you specify its location Otherwise you specify the path to the server s home directory or possibly a bin directory In some case you may be required to provide more information such as a d
465. the following options e Continue the configuration process by specifying advanced JProbe settings Click Next and continue on step 7 e Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 e Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 7 Optional Specify the JProbe advanced options For more information see Specifying Advanced JProbe Options on page 46 In this screen you have the following options e Continue the configuration process by reviewing the settings Click Next and continue on step 8 e Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 e Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 8 Optional Review the settings in the configuration summary screen For more information see Reviewing the Configuration Summary on page 49 The configuration is now complete 9 Save the configuration settings and close the configuration wizard Click Save amp Close 10 Type in a directory and file name for the
466. the path to the nearest interesting dominator Click Find Next Interesting Dominator The tree expands and displays the nearest interesting dominator Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden 196 JProbe User Guide Task Resize column widths Action To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit Response The columns are resized Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted Export the table to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Exploring the Memory Source View The Source view displays source code The source code contains line numbers taken from the line data table that was created when you compiled your code This section contains the following in
467. the trigger action when the action on the n th execution the first time the method is trigger action is performed of the method and then perform the trigger action on every m th execution of the method Stop performing the trigger action when it has been performed x times or when the application ends The values of n m and x are choose n m and x from the drop down list or select the When the application ends check box instead of x executed n minutes after program start and then perform the trigger action the first time the method is executed after every m minutes Stop performing the trigger action when x minutes has elapsed since the trigger first fired or when the application ends The values of n m and x are choose n m and x from the drop down list or select the When the application ends check box instead of x Refining Data Collection and Presentation Changing Trigger Execution Order When two triggers on the same method occur on the same event the first one in the table is executed first Otherwise trigger order in the table is not important 429 Setting Triggers Task Move a trigger Action Click the button at the left of the trigger s row and drag the trigger to the new location Response The trigger is inserted where you drop it Insert an empty row between two triggers Right click the button at the left of the trigger s row and choose Inse
468. the units in the Edit Options dialog e For graphs that you want to print in black and white ensure that the contrast among colors is strong enough Tip Blue with light gray is a good choice for printing to black and white 3 Click Print 3 on the toolbar or click File gt Print 4 If this is the first time printing from JProbe you are prompted to enter the location of the Adobe Reader executable 5 For tables you can choose the rows you want to print 6 Specify the header and or footer Enter text or variables in the Left Center and or Right fields depending on where you want the text to appear in the output You can use the following variables Variable Date Description The date and time that the report was created Class The class name of the application Application The fully qualified class name of the application Page The page number of the report 7 To have components printed on separate pages select the Start Each Component On A New Page check box 442 JProbe User Guide Adobe Reader Executable Location Adobe Acrobat 7 0 Acrobat Acrobat exel Rows to Print All O First Header Left Center Right Footer Left Report Date Date Center Right Page 7 Start Each Component On A New Page 8 Click Print Adobe Reader opens with the view results displayed You can now save and print the results from Adobe Reader E
469. the upper pane For example by summing up Exploring Memory Analysis Views 151 Exploring the Dominating Instances View the Heap Count rows in the lower pane you get a value that corresponds to the Heap Count values in the upper pane Note The Keep Alive Size values are estimates so the sum of the values in the lower pane differs from those in the upper pane The status bar displays the most common icons and short descriptions of what they represent In the Dominating Instances view you can look at the lower pane and quickly see that a number of the instances are pinned by a particular class The middle pane offers a breakdown of this information using categories so you can see which category of classes are holding on to the most of your starting set of instances A good approach is to search for the highest count or keep alive size and focus on those instances by asking Why Are These Live to repeat the process or by selecting the appropriate bar from the chart in the middle pane Once you have reduced the set of instances to a few that seem to be responsible for pinning the majority of memory it is worth opening the Instance Detail view You can drill down in a category by clicking either a section in the chart or the corresponding category in the legend As you drill down to more detailed information the chart and table refresh to show only the instances in that category You can click the back and forward toolbar
470. there is no data is gray e Filtered out catch blocks are gray You can change these colors For more information see Customizing the Colors Used to Display Results on page 392 386 JProbe User Guide Coverage Bar Along the right margin is a navigation aid called the Coverage Bar The length of the Coverage Bar represents all the source code in the file not just the code currently displayed in the view The colored stripes on the bar represent the relative size and position of missed lines of code and missed conditions if available You can click a stripe to navigate to that block of code For more information see Interacting with the Source View on page 388 Missed Condition Column When the snapshot contains data on conditions the Missed Conditions column displays the number of missed outcomes within the context of the total number of possible outcomes For example if one out of two conditions were missed the value 1 of 2 is displayed To understand how JProbe determines the total number of conditions consider the following if statement if a amp amp b amp amp c JProbe calculates two possible outcomes true and false for each of the three subconditions for a total of six possible outcomes 2 2 2 6 Note that the calculation is not based on the combinations that these possible outcomes can generate 2x2x2 8 If the condition outcomes do not match your expectations your compiler may have m
471. thod is started for example by using a trigger on the method itself Performance Session Runtime Tips The following tips will help you to improve session runtime performance Try different supported JVMs The session runtime performance depends in part on the speed at which your Java program runs Modern JVMs such as Java 6 generally improve program performance For more information see the JVM documentation and the JProbe Installation Guide Exit your program before investigating snapshots If you open a snapshot in one of the investigation tools while your program continues to run you may find that the JProbe Console responds more slowly than desired because it is competing for resources with the JProbe Analysis Engine To improve JProbe Console performance exit your program before opening a snapshot or disconnect from the session if you don t want to end your application 352 JProbe User Guide Choose not to monitor the session or monitor it remotely Monitoring the session requires that two JVMs be running the JVM under analysis that executes your program and the JVM running the JProbe Console If you monitor remotely the JVMs are not competing for resources on the same computer Alternatively you can run a JProbe session without the JProbe Console Instead of interacting with the session through the JProbe Console you use triggers to collect data and take snapshots After the session ends you can open the snapsho
472. thods in the report is controlled by the Number of Rows to Report option in the Create Performance Report dialog box Report the top 10 is the default option Create Performance Report Output File C JProbe Reports Polynomial Number of Rows to Report After sorting report top 10 v Generate File Cancel _ Preview _ Help Runtime Summary Report This report contains the content of the Memory Pools and GC Data tabs in the Runtime Summary view Memory Pools Pool Used Mb Committed Mb Available Mb Maximum Mb Total 88 18 7 939 167 3 Heap Subtotal 0 6 5 2 6 66 7 Eden Space 0 0 0 9 og 42 Surwor Space 0 0 0 1 0 1 0 5 Tenured Gen 0 6 4 2 3 6 62 0 Non Heap Subtotal 8 2 13 5 53 100 7 Code Cache 09 0 9 00 33 6 Pern Gen 7 3 12 6 5 3 67 1 Garbage Collection Overhead 0 0 Normal GC Data Colector Count Time ms Cumulative Count Cumulative Time ms Copy 0 0 00 3 21 00 MarkSweepCornpact 0 0 00 2 143 00 If you select the Include Thread Information check box in the Runtime Summary Report dialog box this report also includes the states of all threads currently displayed in the Threads tab Create Runtime Summary Report Output File C JProbe workspace reports DinerReport J Include Thread Information Reporting Performance Analysis Results Summary Report 345 The list includes the identity of any locks these threads are blocked on but no category entries are reported
473. tigate By options 140 JProbe User Guide If you want to Select this Investigate By option See recorded data in subsequent views Recorded Count or Recorded Memory See heap data in subsequent views Heap Count or Heap Memory See garbage data in subsequent views Dead Count or Dead Memory Investigate traces for dead objects Dead Count or Dead Memory and go to the Call Traces or Merged Allocation Points view See instances created during your session Recorded Count or Recorded Memory in the Instance Detail view and go to the Instance Detail view Consider all instances in the heap Heap Count or Heap Memory instead of just those watched while recording Some views only display certain types of data For example the Merged Allocation Points and Call Traces views display only recorded data or garbage collection data You will not be able to access these views when Investigate By is set to Heap Count or Heap Memory or when the chosen class does not have any recorded or garbage instances The bar chart in the Instances view illustrates the percentage of instances in each of the categories that are listed in the legend to the left of the chart Hovering over a segment of the chart displays a popup showing the number of instances in or memory used by that category You can drill down in a category by clicking either a section in the chart or the corresponding category in the legend As you drill down to more detailed info
474. tion name and select Edit or Click Configurations gt Edit lt Configuration_Name gt on the menu bar The lt Settings_File_Name gt tab is placed in editing mode 3 Edit the settings on any of the tabs as necessary 4 Click Save Deleting a Configuration To avoid confusion you should consider deleting configurations that you no longer use To delete a configuration 1 Select the configuration in the Configurations Navigator 2 Inthe lt Settings_File_Name gt tab click Delete or Right click the configuration name and select Delete or Click Configurations gt Delete lt Configuration_Name gt on the menu bar 3 Click Yes in the Delete Confirmation dialog box The configuration is deleted from the Configurations Navigator Creating a Configuration from a Settings File You can create a configuration from a settings file For example a colleague may give you a settings file to use as a starting point for your own analysis During the load process if the settings file contains invalid options or if there are naming conflicts JProbe lists the issues in a message box JProbe then attempts to create a configuration using the valid values in the settings file To create a configuration from a settings file 1 Right click in the Configurations Navigator and select Add Settings File 62 JProbe User Guide The Load JProbe Settings File dialog box appears 2 Navigate to and select the settings file and click Open JProbe inspect
475. to connect JProbe to your application JProbe begins to collect data on your program and displays it in the Runtime Summary view 4 Inthe Runtime Summary view click Set Recording Level x on the toolbar The Set Recording Level dialog box appears t Set Recording Level To set the recording level move the slider The position of the slider indicates the tradeoff between the amount of data collected and the impact on your application s performance Less Data Faster Data Recording Off Public Methods No Encapsulation Collect data only for public methods that match your filters No Encapsulation Collect data only for the methods that match your filters Grouped Encapsulation Collect data for your methods and combined data for the methods they call Full Encapsulation Collect data For your methods and the methods they call More Data Slower 5 If data recording is off turn it on by moving the slider down for details see Setting the Recording Level on page 236 and click OK Note Unless the Data Recording is off a new recording session always begin after you click OK even if you did not change the level 6 Exercise your program Running a Performance Analysis 239 Running an Analysis Session 7 If desired take one or more snapshots by clicking Take Snapshot 3 on the toolbar 8 When the program has completed click Set Recording Level Recording to off and click OK set the Data The data
476. top of the bar to move to the previous or next missed code For more information see Exploring the Coverage Source View on page 384 6 Determine if the missed lines of code in the method require testing and adjust your test cases You may need to consult the software developer in charge of the code If so you can print a report of your results 7 If you want to review all missed code in this source file click the Next Stripe arrow on the Coverage Bar The view scrolls to show the next missed line 8 To investigate a missed method in another source file select the Snapshot Browser tab and repeat step 4 to step 7 9 When you are done rerun the session with the improved test cases Next Steps As your program grows and evolves your test suites need to keep pace with the changes Changes are not always communicated to QA analysts After you have achieved a satisfactory level of coverage you can continue to use the JProbe Coverage analysis tool regularly If your percentage of missed methods or missed lines of code 366 JProbe User Guide goes up it is an indication that development has added new code that is not covered by your test cases 1 Learning More About Coverage Options In Running a Coverage Analysis on page 359 the default options were used You can exercise more control over your analysis by experimenting with other options This chapter presents how to fine tune the Coverage analysis tool and contain
477. tract lt license_number gt lt license_file gt e UNIX jplicense extract lt license_number gt lt license_file gt where Jicense_number is the license to extract and license_file is the fully qualified path to the file where you want to save the extracted license To get a list of license numbers use jplicense list You can now use JProbe Per Seat Licensing Seat licensing allows one individual user to use JProbe for a designated platform This option is not covered in this guide For more information contact your Sales representative JProbe License Options 469 Concurrent Licensing Concurrent Licensing Concurrent licensing offers flexibility in where JProbe is installed and who may use the product Instead of having one node locked license per developer you purchase a number of licenses that are shared by the entire team via the Quest Software License Server Note Concurrent licensing is not available with evaluation versions of JProbe When you launch JProbe the product checks to see if a license is available from the License Server If so you can use the product if not you need to try again later If developers consistently have trouble obtaining a license you can increase the number of licenses by contacting Quest Software For more information see Using Concurrent Licenses on page 469 Using Concurrent Licenses Concurrent licensing means that multiple people can start simultaneous JProbe analysis
478. tric that you are interested in Allocations by Method Count Name Method Count Method Memory Bytes Cumulative Count Cumulative Memory Bytes javax swing 48 936 155 7 040 java awt 25 512 158 7 120 java lang 24 904 26 960 demos memory leakexample 14 3 528 126 5 880 javax swing event 14 448 14 448 java beans 7 112 14 224 javax swing plaf basic 7 168 132 5 976 java lang reflect 7 112 7 112 sun java2d d3d 6 304 13 536 sun java2d 4 160 8 280 java util 3 88 3 88 sun swing 2 64 18 720 sun awt image 2 264 16 848 java lang ref 1 32 1 32 sun java2d pipe 1 32 1 32 javax swing plaf metal 1 24 26 912 java security 1 24 1 24 sun awt windows 1 32 1 32 sun awt 0 0 8 152 java awt event 0 0 23 1 064 java awt image 0 0 3 288 sun awt im 0 0 1 32 Allocations Detail Report This report contains the content of the table in the Allocations Detail view Before generating this report sort the table by the metric that you are interested in Reporting Memory Analysis Results Instance Detail Report Allocations Detail by Count Name Count Memory Bytes javax swing event EventListenerList 14 224 javax swing AbstractButton Handler 7 112 javax swing DefaultButtonModel 7 280 javax swing InputMap 7 112 javax swing ActionMap 7 112 javax swing event ChangeEvent 6 96 Instance Detail Report 207 This report contains the contents of the upper table in the Instance Detail view Before generating this report sort the table by the metric that y
479. trics include all invocations of the method on all threads during the execution of your program Basic Metrics The following table presents the basic metric types Basic Metrics Description Number of Calls Number of times the method was invoked Method Time Amount of time spent executing the method excluding time spent in methods it called Number of Catches Number of times the method catches an exception Number of Exception Number of times the method exits by exception Exits Catch and Exception Counts JProbe collects and displays method level information on how many exceptions are caught by instrumented methods and how many times those methods were exited via the exception mechanism This information displays in several Performance views including the Methods Method Detail Snapshot Difference and Performance Report Running a Performance Analysis 241 Running an Analysis Session Metrics and Threaded Programs In multithreaded programs data for a method includes all invocations of that method on all threads For example if thread A invokes methoda four times and thread B invokes the same method six times the Number of Calls metric for methodA is ten 4 6 10 While it may seem obvious totaling time on all threads can lead to times for the program that do not match the time you subjectively experienced Consider the following example where methodA is executed by two threads and time is measured as elap
480. ts in the JProbe Console Set filters to reduce the amount of data collected Filters reduce overhead and improve session performance significantly The default filter setting collects data on methods in your package and encapsulates data for the methods contained in all other packages If you were to collect data on all methods the session would run very slowly and the data collected would be of little value in tracking down performance bottlenecks As you confirm which of your methods are working well you can encapsulate those methods too For more information see Encapsulating Data on Methods in the JProbe User Guide Only collect line data on problem methods Only use the Line Level filter on methods that you have identified as performance bottlenecks Collecting data line by line causes the session to run more slowly For more information see Collecting Data on Lines of Code in the JProbe User Guide Reduce the number of nodes in the Call Graph Displaying a large or complex Call Graph can take longer than desired To improve the responsiveness of the Call Graph use some of the following techniques to reduce the number of methods that need to be maintained e use data collection filters e exclude threads e prune subtrees e isolate a subtree Troubleshooting Performance Analysis Results 353 Performance Help Messages Performance Help Messages When you try to execute an action that is not enabled the JProbe Co
481. ture in the cell For more information see Syntax for Filters on page 41 Set the Action to Exclude Repeat to exclude additional code Order is important Filters lower in the table override filters defined higher in the table If you need to rearrange the filters click and hold the row button Configuring JProbe to Analyze Your Application 44 Specifying Your Code located at the beginning of the row for the filter you want to move and drag it to its new location My Application Filters package class method Action _ Ce include fg v v demos gamepack framework image GameImage exclude v V demos gamepack framework faults Deadlock exclude v demos gamepack framework loadtest LoadTestServlet exclude v To enable all application filters on the My Application Filters table e Select the check box located on the left side of the My Application Filters table title The check boxes corresponding to all application filters on the list are selected automatically To disable all filters on the list click the My Application Filters check box again To delete a filter from the My Application Filters table 1 Select the row containing the filter you want to delete from the list 2 Do one of the following e Right click the row button located at the beginning of the row and click Delete e Press Delete on the keyboard The selected filter is deleted f
482. uDeraccessimgt Java lo Consalle i Java securty SaskPermssionC olection sun java2d Joops ComposkeType sunma Metalnder sun font PieFonStrike double C10 sun nio c5 M51254 necoder prva sveng BufferStrategyPantstanager javax sring UaDef auts sun maniagemerk Mernor yManager iret sun font attrbutevalues JavIK srang Componentingumap Sun ant Win32GraphicsEnvircemert java ubl Hashtable Emptyiter ator soun font PtaysicalFort java ant ComponentOnent ation sun java2d oops GraghicePrintiveMgr 2 sun javaza nutsurfeceData java Plelnput3reom Understanding the Memory Difference View 4 10 400 160 s C 104 p 160 456 168 104 ececeoeseescecee oe see e sooo ssc eo om 11 074 20 280 o 168 6 976 rd 18 The Memory Difference view is similar to the Instances view except that it displays differences between one snapshot and another Values In the table a negative value means that you improved memory use fewer instances were created or less memory was used in the compared snapshot A positive value means memory use worsened since the baseline snapshot more instances of the class were created in the compared snapshot A zero value indicates no change Across the top of the table is a totals line which shows the overall change the underlying data set Note If you have used the Filter Classes option some of the classes may not be visible 200 JProbe User Guide Per
483. uest com for more information Introduction to This Guide 21 About Quest Software Inc Contacting Quest Software Email info quest com Mail Quest Software Inc World Headquarters 5 Polaris Way Aliso Viejo CA 92656 USA Web site www quest com Refer to our web site for regional and international office information Contacting Quest Support Quest Support is available to customers who have a trial version of a Quest product or who have purchased a commercial version and have a valid maintenance contract Quest Support provides around the clock coverage with SupportLink our web self service Visit SupportLink at http support quest com From SupportLink you can do the following e Quickly find thousands of solutions Knowledgebase articles documents e Download patches and upgrades e Seek help from a Support engineer e Log and update your case and check its status View the Global Support Guide for a detailed explanation of support programs online services contact information and policy and procedures The guide is available at http support quest com pdfs Global Support Guide pdf Quest Communities Get the latest product information find helpful resources and join a discussion with the JProbe Quest team and other community members Join the JProbe community at http jprobe inside quest com 22 JProbe User Guide 1 Configuring JProbe to Analyze Your Application This chapter prov
484. uide e JBoss v4 2 x on page 86 e JBoss v5 x on page 87 e Open Source GlassFish v2 UR on page 87 e Open Source GlassFish v3 on page 88 e Open Source Java Service Wrapper v3 3 on page 89 e Oracle AS 10g v10 1 3 0 or below on page 90 e Oracle AS 10g v10 1 3 1 or above on page 91 e Oracle WebLogic v10 3 on page 92 e Oracle WebLogic Server 11g R1 v10 3 1 on page 92 e Oracle OC4J 10g v10 0 x on page 93 e Oracle OC4J 10g v10 1 2 or below on page 93 e Oracle OC4J 10g v10 1 3 on page 93 e SpringSource dm Server v1 0 2 on page 94 e SpringSource dm Server v2 0 on page 94 Apache Geronimo v1 0 You run the JProbe generated startup script in the same way that you run your Geronimo startup script Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Usage lt configuration_startup_script gt run start geronimo_args where e run Start Geronimo with JProbe in the current window Starting Your Application with JProbe Attached 77 Runtime Notes f
485. unt Dead Memory Total 27 15 947 1 256 1 181 760 1 181 760 8 324 368 520 char 1 3 991 88 245 408 245 408 424 21 408 java lang Class o 1 323 a 572 584 854 632 o a java util HashMap Entry o 994 a 23 856 94 096 164 3 936 java util HashtablegEntry o 821 0 19 704 40 568 0 o java lang Object 2 406 112 21 592 24 528 585 13 536 java lang Integer o 351 o 5 616 5 616 10 160 java lang Short o 256 a 4 096 4 096 o a sun java2d loops Blit o 199 o 7 960 7 960 o o sun java2d loops ScaledBlit o 163 0 6 520 6 520 o o java lang String o 155 a 7 320 16 848 o a java lang Character o 129 a 2 064 2 064 o a java util HashMapgentry 2 11 160 14 360 46 376 46 3 680 java util HashMap 2 109 80 4 360 49 936 46 1 840 sun java2d loops MaskBlit o 88 0 3 520 3 520 o o java lang Object i 82 8 656 656 10 ao The Investigate By list above the chart allows you to select which metric should be used to determine the top five categories The following options are available e Recorded Count e Recorded Memory e Heap Count e Heap Memory e Dead Count e Dead Memory It is important to keep in mind that the option you select from this list affects not only the Instances view but also the availability of subsequent views If you cannot access a view or you do not see the data you expect to see in a subsequent view it may be due to the Investigate by option that you select in the Instances view The following table lists some examples of how to use the Inves
486. untime Summary r Ca 7 RaT Configure Pools Pools and GC Interval One Minute Classes Loaded 1938 Classes Unloaded 0 Memory Pools GC Data Threads ia aa fi Thread States click on a bar to show threads in that state Number of Threads Runnable Waiting Biblocked Hother 0 1 2 3 Md 5 6 7 8 9 10 Total Threads 25 Last Updated 11 24 37 Filter Threads 25 25 show all _ Update Thread Details Detect Deadlock Neme Philosopher 1 State BLOCKED on demos performance diners ChopStick l fdac pi Philosopher 2 Total blocked 2 Total waited 5 Philosopher 3 Philosopher 4 Lock Ower Philosopher 2 Philosopher 5 Philosopher 6 Object Monitors Locked by Thread Philosopher 7 demos per formance diners ChopStick 2431b9 Philosopher 8 Philosopher 9 T ihes Ph ds Senos per formance diners Philosopher run Philosopher java 104 p aT A S java lang Thread run Unknown Source java UI Threads 5E Java VM Threads Se IMX Threads Investigating Performance Deadlocks After you execute the procedure described in section Finding Deadlocks the lower pane in the Threads tab is updated to display only the threads that caused the deadlock s These threads are displayed in the graph as Blocked To view details about a particular thread involved in the deadlock situation Click the thread name in the threads list The following information is displayed in the
487. update the org osgi framework bootdelegation property as follows org osgi framework bootdelegation org w3c org xml sun com jprobe Note An error message is displayed during the integration process when no com jprobe filter strings exist in the java6 server profile and java5 server profile files You run the JProbe generated startup script in the same way that you run your startup script startup bat or startup sh JProbe requires the same environment variables such as JAVA_HOME and the same command line arguments Important It is recommended that you use the default script directory that is the script directory shown in the startup script field during the integration process or use the original startup script directory to save the JProbe startup script file Failing to comply to this recommendation may result in the JProbe startup script not working correctly SpringSource dm Server v2 0 The SpringSource dm Server requires the filter string com jprobe to be added to the org osgi framework bootdelegation property in the SERVER_HOMEMib java6 server profile file You must manually add these filters into the file before integrating the SpringSource dm Server with JProbe For example you can update the org osgi framework bootdelegation property as follows Starting Your Application with JProbe Attached 95 What to Do When the Console Does Not Connect org osgi framework bootdelegation
488. updated every four seconds Welcome W Runtime Summary Configure Pools Pools and GC Intervali One Minute Classes Loaded 1939 Classes Unloaded O Memory Pools GC Data Threads Exploring Performance Analysis Views 277 Exploring the Runtime Summary View zyrt 2a a fit Number of Threads Runnable Waiting Bislocked Hother Thread States click on a bar to show threads in that state 1 2 3 4 Total Threads 20 5 6 7 8 9 10 Last Updated 15 16 32 Filter Threads 20 20 Show All Detect Deadlock Philosopher 1 Philosopher 2 Philosopher 3 Philosopher 4 Philosopher 5 Java RMI Threads 5H Java UI Threads 52 Java VM Threads Z MX Threads The graph displays how many threads are in a given state at a certain time The threads are grouped as presented in the following table Group Thread State Thread Description Runnable RUNNABLE A thread executing in the Java virtual machine Waiting TIMED_WAITING A thread that is waiting for another thread to perform an action for up to a specified waiting time WAITING A thread that is waiting indefinitely for another thread to perform a particular action Blocked BLOCKED A thread that is blocked waiting for a monitor lock Other NEW A thread that has not started yet TERMINATED A thread that has exited 278 JProbe User Guide Clicking on a bar or a line in the legend redu
489. uration and click Next For more information see Defining a JProbe Configuration on page 29 Specify the main class of your application working directory the class path and any application arguments and click Next For more information see Specifying Your Java Application on page 31 Specify a supported JVM plus any Java options you normally use and click Next For more information see Specifying a Java Virtual Machine on page 36 Specify application filters enable JProbe to create a category containing the code set to include and click Next For more information see Specifying Your Code on page 38 Select the type of analysis you want to run For more information see Specifying a JProbe Analysis Type on page 45 In this screen you have the following options e Continue the configuration process by specifying advanced JProbe settings Click Next and continue on step 9 e Save the configuration settings and close the configuration wizard Click Save amp Close For more information see Saving the Configuration on page 50 e Save the configuration settings and integrate your application to run with JProbe Click Save amp Integrate For more information see Saving the Configuration and Integrating on page 55 Optional Specify the JProbe advanced options Configuring JProbe to Analyze Your Application 29 Defining a JProbe Configuration For more information see Spe
490. ut the JProbe version the application and the environment in which they run For a snapshot it also contains the snapshot name and file size Actual Contains the options that were passed to the JProbe Analysis Engine expressed as command line JProbe options These may be different than the options that were requested in the configuration Requested Contains the options that were set in the configuration expressed as command line JProbe options JVM Contains information about the Java executable running the application To view the execution log 1 Do one of the following e Inthe Runtime view click Runtime Execution Log gt on the toolbar e In the Snapshots Navigator right click a snapshot and select Execution Log PS bj Communicating Results 445 Viewing the Session History A Execution Log General Actual Requested JVM Setting Value JProbe Version 8 0 0 Program Polynomial Program Start May 2 2008 6 53 29 AM Use Case Time May 2 2008 6 53 29 AM Machine Name abc123 OJS Name Windows xP OfS Version 5 1 2 Review the information 3 Click Close Tip Errors and warnings that have occurred before users connected to the JProbe Analysis Engine are displayed on the system messages area in the runtime views not on the Execution log Viewing the Session History When running a session you can access the Session History from the Runtime Session views The Sess
491. view by class 1 Type a pattern or comma separated list of patterns in the Filter Classes text box 2 Press Enter to apply the filter The table refreshes listing only the classes that meet the criteria you entered The number of classes total classes is displayed beside the text field Note that the chart is not affected Note The filters you enter in this field are saved and can be used later by clicking the down arrow and choosing them from the drop down list To filter the Instances view using the Set Filters dialog 1 Click Set Filters 7 The Set Filters dialog box opens 142 JProbe User Guide E Sot Filtors Set filters by adding categories to the fiter ist You can also type categories and patterns into the ist Categones Add the View s Current Piters to the New Piters List Add Current Piters CE System Code 2 D P aty Framencrts New Filters CE My Code 2 Add filters to the New Filters list in one or more of the following ways e Click a line in the list to highlight it and type a pattern e Click in a line in the list to highlight it then click the browse button to navigate to a class or method and double click it e Click Add Current Filters to add the filters that are currently in effect in the Instances view e Navigate to a class or method in the Categories pane and click Include 3 If you want to change the action for a filter click in the Action column and choose the Include or Exclude
492. w Task Display non heap data Action Click the Configure Pools button to open the Select Memory Pools dialog box then select the Non Heap option Note This action may not be available for all JVMs Response The graphs refresh to display the non heap data Information on other data structures that are occupying memory such as the permanent generation may be included if the JVM supports it Change the data interval for the memory pools and garbage collections Select the appropriate option from the Pools and GC Interval list The Memory Pools and Garbage Collection graphs change to the selected interval View thread details In the Threads tab click the Update Thread Details button or click a thread name Note The button is disabled when no threads are selected or a category is selected JProbe displays the details for the currently selected thread Detect a deadlock Click the Detect Deadlock button If a deadlock is detected the threads list displays the threads involved in the deadlock If no deadlock is detected a message is displayed to that effect Save the chart as a PNG or JPEG file Right click in the chart and select Save Chart As then select the graphic file type PNG or JPEG in which you want to save the chart In the Export Image dialog box specify the export parameters and a location to save the file The file is saved to the locatio
493. w on page 141 Show hide table columns Right click the table and select Show Hide Columns In the Show Hide Columns dialog box select the columns to show and click OK The selected columns are displayed The rest are hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All Columns to Fit The columns are resized Rearrange columns Drag a column by its heading and drop it in a new location The column appears in its new location Sort the table by a metric Click the column heading for that metric To reverse the sort order click the heading again The table is sorted Export table data to CSV Right click in the table and select Export to CSV Type a file name and click Save The file contains the table data separated by a comma It may also contain headers For more information see Changing How JProbe Exports to a CSV File on page 455 Exploring Memory Analysis Views 163 Exploring the Allocations Detail View Create a report for Click Export Data to The Create Allocations this view HTML a Report dialog box appears For more information see Allocations Report on page 206 Exploring the Allocations Detail View Th
494. w Configuration Wizard Java Application Specify Your Java Application Tell us about your Java application Main Class Execute a class Execute a Jar file with manifest Main Class demos memory leakexample LeakExample Working Directory Quest_Software JProbe_8 3 Application Arguments Classpath Required Field Main Class and Working Directory You can select the main class or point to a JAR file containing the main class To be able to select a JAR file the JAR file must contain a manifest Manifest files are generated automatically during the archive process or you can specify your own manifest file For more information see the Sun Microsystems Web site By default the Main Class dialog box available from the Browse button lists all the classes To modify this default behavior use the Options tab in the Main Class dialog box to select the type of files to be filtered out The following options are available e Show only classes with main method to include only the main classes e Search JAR manifest for main class to include JAR files that contain a main class Both options are by default disabled They are global options and they are remembered for the next time you open the file chooser Important If you have a lot of files in your working directory the file chooser needs a lot of time to filter out files for the first time In this case it is recommended to initial
495. whose name starts with Pre matches both Pkg Prepare m and Pkg Pretend C m All methods in classes whose name ends with y and either starts with s or belongs to a subpackage that starts with s matches both java rmi server RMISocketFactory com quest say and java security cert X509CRLEntry All methods in the class bar in package foo This is likely not what you wanted See the next example 436 JProbe User Guide If you type this Canonical formis And filters are applied to this foo bar foo bar All methods in all classes in f00 bar or any of its subpackages foo pat no change Methods whose names start with t and end String t e in e in foo bar String matches both toUpperCase and toLowerCase Adding Folders To add a folder 1 Select Tools gt Manage Categories 2 Select a folder and click New Folder 3 Enter a name in the Name field 4 Click Apply 5 Repeat to create additional folders 6 Click OK Editing Categories To edit a category 1 2 3 Select Tools gt Manage Categories Select a category and click Edit Make changes to the filter list Note A Category created automatically by JProbe should not be renamed or at least a folder with the original name should be preserved otherwise JProbe creates it again Click Apply Click OK Refining Data Collection and Presentation 437 Creating and Managing Categories Renaming Categories or Folders To rename a category or fol
496. wser The Snapshot Browser displays data from the currently chosen snapshot This section contains the following information e Opening the Snapshot Browser e Understanding the Data in the Snapshot Browser e Interacting with the Snapshot Browser Opening the Snapshot Browser The Snapshot Browser opens automatically when your application exits normally and displays the contents of the last snapshot taken You can view other snapshots in the Snapshot Browser To open a different snapshot 1 Inthe Snapshots Navigator double click the snapshot that you want to display Understanding the Data in the Snapshot Browser Data in the Snapshot Browser is presented in terms of misses that is the classes methods lines of code and conditions if the snapshot contains condition coverage data that were missed by your test case The Snapshot Browser is divided into two areas In the upper area called the Application Tree your application s packages and Exploring Coverage Analysis Views 379 Exploring the Snapshot Browser classes are organized in a tree format When you choose a package or class in the tree the Method Detail table displays data for each of its methods ESS amp coverage Session Snapshot Browser 7 aa Name I Missed Classes Missed Methods Y Missed Lines With Line Data Adventure_TestCase1 16 7 W 22 6 E 31 7 B 100 0 5 demos coverage adventure 16 7 E 22 6 E 31 7 B 100 0 Kitchen
497. xample when or where these instances were allocated you must turn on the data recording Enabling the recording for those instances is recommended when tuning the garbage collection and when analyzing over allocation issues The Set Recording Level dialog box presents the following options Select this option Data Recording Off In this situation You are at the initial stage of your investigation when you do not know what is leaking or why This is often enough to at least start your investigation Note Recording should also be Off while your application and or application server is starting unless you need to record the startup behavior Record Allocations For All Instances You want to record only the allocation times of instances which are displayed in the Instance Detail view The Call Traces and Merged Allocation Points views are not available in this case as no trace data is collected Note If you have a very large program if you are using an application server or if you Know what class of objects you want to track consider using Record Data as Specified in the Table Below because Record Allocations For All Instances may have a severe performance and memory impact Record Allocations and Stack Traces For All Instances You want to record allocation times as well as stack traces You can use the Call Traces and Merged Allocation Points views to analyze where instances came from Note If
498. xists for this case e You have attempted to take a snapshot with a heap dump but the JVM failed to provide a valid dump Workaround Try to reduce the load on your application before taking the snapshot if possible Let the application and the JVM become as idle as possible for example for a Web application stop making requests and manually request a garbage collection GC Wait until you see the effect of the GC in the Runtime Heap graph do another manual GC then wait again Take a new snapshot Failures sometimes can be sticky that is once a JVM fails to take a heap dump in arun subsequent attempts may also fail Therefore if you get a failure and the workaround suggested here does not help shut down the application or application server and restart it this time trying the workaround suggestions for every snapshot 216 JProbe User Guide No Stack Trace Data You can view stack traces only if your snapshot contains stack trace data The type of data collected during a session is defined by setting the Memory Recording Level appropriately If you want to collect stack trace data you must set the Memory Recording Level to one of the following options then run a new session to obtain a new snapshot Record Allocations and Stack Traces For All Instances e Record Allocations Stack Traces and Garbage Data For All Instances e Record Data as Specified in the Table Below define filters for problem classes an
499. xporting Table Data to a CSV File When exporting to CSV comma separated values all visible data is included in the output even if you need to scroll to see it on screen You can set up the view the way you want the data to be exported In the JProbe preferences you can select whether or not headers are included in the exported data For more information see Changing How JProbe Exports to a CSV File on page 455 To export a table of data 1 Open the view 2 Set up the view You can set the table sort order the columns shown the order of those columns and view filters if available You can also set the units in the Edit Options dialog 3 To export the data right click in the table and choose Export to CSV 4 Specify a location and file name 5 Click Save Communicating Results 443 Exporting Runtime Chart Data to a CSV File Exporting Runtime Chart Data to a CSV File When exporting a Memory or Performance runtime chart to CSV comma separated values you can include heap data non heap data or both The heap data includes how memory is used in the heap for example you might see Eden Space Survivor Space and Tenured Gen values for each timestamp The non heap data includes other kinds of data like Code Cache and Perm Gen for each timestamp Timestamps reflect clock time not the elapsed time displayed in the chart and are reported at five second intervals Note The Coverage Runtime view does not include a
500. y a bug but one that is hard to track down If you approach the problem as a loitering object you may find it more quickly Maintenance of singleton objects can be made easier by encapsulating state transitions in a single method rather than scattering changes of state throughout your code Stalled Stack References Long running methods and stalled threads can pin objects in the heap until the method or thread returns For example consider the StalledStack class application in the demos directory The demo reads an XML file using DOM parses out Java file names and then performs a long running dummy process on the parsed file names without freeing the memory taken up by the DOM tree The memory will eventually be freed when the long running process completes but in the meantime memory that could be used for the process is unavailable Setting the Recording Level Before you start a Memory analysis you need to decide what kind of data you want to see The data recorded during the session provides you information about Where did this object come from Knowing where an instance came from can often help you track down a cause and resolution for a memory leak Running a Memory Analysis 111 Finding Memory Leaks When dealing with a memory leak you would be mostly interested in Why is this object live information that can be determined by examining the referrers in the heap For additional information about the instances for e
501. y be incorrect Learning More About Coverage Options 371 Analysis Options for Coverage Triggers for Coverage The Analysis Type gt Automation tab allows you to set triggers and data recording for your analysis Triggers Triggers provide a way to tie JProbe actions to your program s execution You can have JProbe perform an action when your program enters or exits a method of your choice With advanced trigger controls you can also specify when to start executing the trigger actions and how long to execute them Triggers are particularly useful for automated sessions Summary Analysis Type Java Application My Code Advanced JProbe Options JVM Analysis Type Memory Analysis Identify and resolve Memory leaks and object cycling Performance Analysis Identify and resolve Bottlenecks and deadlocks Coverage Analysis Identify Unexecuted line of code P General JP Filters Automation Trigger On Method Event Action Parameter Add Trigger Please enter a method to set a trigger on S B Initial Recording at JVM Start Data Recording Off Data Recording On Triggers are defined in the same way for all the analysis tools although the trigger actions are somewhat different This section describes the trigger actions for a Coverage analysis To learn how to create triggers see Setting Triggers on page 425 The table below describes the trigger actions for Coverage Trigger
502. y using the options provided and click Add to display them in the text field or Choose Type a Query and type your query in the Query field 4 When you have built or typed your query you can save it by giving it a name and clicking Save 5 Click Submit to execute the query All the instances that do not meet the parameters of the query are removed from the list Interacting with the Instance Detail View You can select a new instance and review its tree of referrers or references In the tree you can expand and collapse branches At any time you can view source code for a class from an instance or allocating method JProbe User Guide The following table summarizes the tasks that you can perform Task See Referrer Reference data for an instance Action In the instance list click an instance row Response The Referrers References tab displays data for the instance See Dominators for an instance In the instance list click an instance row The Dominators tab displays data for the instance Expand or collapse a tree or branch Double click the Referrer or Reference instance or click the expand symbol If collapsed the branch is expanded If expanded the branch is collapsed View the source code for an allocating method Select an instance right click and select Show Allocated At Source Note _ If this menu item is disabled JProbe cannot associate a source file wit
503. you choose for the view Data is always sorted in descending order HTML reports are created using valid XHTML 1 0 which means that they can be parsed and manipulated by XML tools Note You can generate many HTML reports using the command line tool jpreport For more information see the JProbe Reference Guide To create a report 1 From a view that supports reports click E Export Data to HTML 2 Specify the fully qualified path to a file in the Output File field 3 Set report options For more information see e Reporting Memory Analysis Results on page 203 e Reporting Performance Analysis Results on page 343 e Reporting Coverage Analysis Results on page 399 4 To generate and display a sample of the report click Preview 5 To generate and save the report click Generate File Printing Data to a PDF File Communicating Results 441 Printing Data to a PDF File When printing to PDF the output is WYSIWYG You can set up the view the way you want to see it before printing All visible data in the view is included in the output even if you need to scroll to see it on screen Adobe Reader is launched automatically from JProbe You can view save and print the PDF from Adobe Reader To print the contents of a view to a PDF 1 Open the view 2 Set up the view e For tables you can set the table sort order the columns shown the order of those columns and view filters if available You can also set
504. you how to configure JProbe to run your application start your application from JProbe run JProbe analysis sessions analyze your results and communicate those results to others It also includes information on troubleshooting strategies for using JProbe and other general help This document is intended for Java developers who want to analyze the memory usage and performance of their code or find out how well their test cases cover their code The JProbe User Guide is organized as follows Chapter 1 Configuring JProbe to Analyze Your Application provides detailed information about the process of configuring JProbe to run your Java EE or Java SE application Chapter 2 Starting Your Application with JProbe Attached explains how to run your application under JProbe and connect a JProbe Console to the JProbe Analysis Engine It also presents how to attach a JProbe Console to an application started without JProbe Introduction to This Guide 17 About This Guide Chapter 3 Strategies for Memory Analysis provides recommendations for how to use the JProbe Memory analysis tool in your development environment Chapter 4 Running a Memory Analysis describes how to run an analysis session locate and investigate loitering objects and presents some techniques for solving memory problems Chapter 5 Learning More About Memory Options presents how to fine tune the Memory analysis tool and how to analyze a heap dump Chapter 6 Explor
505. ystem classes third party packages and frameworks is encapsulated For more information see Encapsulating Data on Methods on page 250 Line Level Use the Line Level detail filter only after you identify hotspots and then only when you need a line by line breakdown of how time is spend in a method Collecting line level detail takes more time and produces larger snapshots For more information see Collecting Data on Lines of Code on page 255 Native By default time spent in native methods is attributed to the method that calls it Use the Native filter to identify native methods in your code The time for these methods is then tracked separately from its calling method You do not need to use Native filters if your JVM is Java 6 or newer Native filters are only required for older JVMs where native methods cannot be instrumented directly Also you do not need to identify native methods in third party packages and frameworks Note Do not use Method Level filters for any native methods If you do use Method Level filters for any native methods JProbe warns you and suggests a native filter If you ignore the warning the data JProbe collects may be incorrect Caution Filter order is important When two entries refer to the same code the entry lower in the table dominates To set data collection filters 1 Click Tools gt Create Edit Settings to open the JProbe Create Edit Settings dialog box Runnin
506. zer for statement Statements table right click and Oracle opens with the select Tune SQL Statement Note This option is available only if you have already installed Quest s SQL Optimizer for Oracle on the machine that will run the JProbe Console SQL text already inserted You can run the tool to optimize the SQL statement View grouped SQL statements Note When a row in the table represents more than one SQL statement the SQL column displays only an asterisk Select the asterisk row in the SQL Statements table right click and select View SQL Statements The SQL Statements dialog box opens displaying all the statements represented by the selected row Show hide table columns Right click in the column and select Show Hide Columns Select the columns to show and click OK The selected columns are displayed The rest are hidden 296 JProbe User Guide Task Hide the column currently selected Action Right click in the column and select Hide Column lt column name gt Response The column currently selected that is the column based on which the table is sorted is hidden Resize column widths To set the column width drag the edge of the column heading left or right To see all data in a column right click in the column and select Size lt column name gt to Fit To fit all columns in the viewing area right click and select Size All

JProbe User Guide

Contents

Download Pdf Manuals

Related Search

Related Contents