PARMON User Manual
Contents
1. PARMON User Manual Chapter 1 Introduction Advanced Computing applications of current interest consume computing power of the order of several GigaFLOPS Vector Processing Systems Massively Parallel Systems and Cluster systems are the means of achieving such orders of computing performance Cluster computing systems consist of relatively inexpensive workstations connected by means of a network Cluster computing systems apart from being economical also offer a high degree of performance scaling Systems Managers of large cluster systems often find it difficult to monitor such systems as it requires the manager to initiate sessions on each such workstation in the cluster that needs to be monitored In addition lengthy commands with specific options need to be remembered to perform such tasks Though several monitoring tools are available they do not cater to the needs of cluster systems PARMON is a tool that has been designed to provide Managers with a graphical interface and allows monitoring of the systems in the cluster from a single workstation 1 1 About PARMON PARMON is meant for managers administrators of Cluster computing systems However it can also be utilised by system managers of single workstations In order to achieve cross platform compatibility the tool has been divided into two parts the server program and the client program The server program runs on the nodes in the cluster that need to be Intr2. Output Memory button Maximum available memory Available Free memory Reserved swap size Free swap size Swap size available for swapping purposes Available resident memory Physical memory size Physical page size Network button Number of network interfaces and their names Figure 37 to Figure 39 list the sample outputs of the logical view of a node Figure 36 Logical View CPU 44 Working with PARMON PARMON User Manual Figure 37 Logical View Disk Figure 38 Logical View Network Working with PARMON 45 PARMON User Manual Figure 39 Logical View Memory Error Message Note The PARMON server must possess superuser rights to display the Logical View of memory The absence of superuser rights causes an error message window similar to that depicted in figure to be popped up 3 4 5 Kernel Data Catalog Kernel Data catalog statistics present managers with a catalog of the resources CPU Disk Memory and Network available on the nodes being monitored To access these statistics select the icon named Kernel Data Catalog from the main screen Figure 40 to Figure 44 depict the appearance of the windows for monitoring the resource utilization of different resources namely the CPU Disk Network and Memory 46 Working with PARMON PARMON User Manual Figure 41 Kernel Data Catalog with drop down list Working with PARMON 47
3. monitored from a single instance of the client program If the server has to be bound to a privileged port 0 1023 then the server program has to be invoked by a privileged user It is recommended a privileged user invoke the server thereby granting the server access to kernel tables The absence of such privileges would render many of the features of the tool unusable 4 Resource amp Setup Requirements PARMON User Manual 2 2 2 Setting up the client Setting up the client consists of the following steps 1 The client distribution consists of a java archive named parmon jar and several GIF images To setup the client copy the java archive to a directory of choice The GIF images should then be copied to a directory named images case sensitive under the directory where the java archive is stored For example if the java archive is stored in a directory named parmon then images should be a sub directory under parmon Also create a directory named setup_files under the directory where the java archive is stored The setup_files directory is used to hold configuration information critical for the functioning of PARMON Set update the following environment variables a CLASSPATH Set the value of this variable to point to the directory of the default java class library classes lib and the java archive parmon jar The format of definition of the value is as follows lt
4. PARMON User Manual Figure 43 Kernel Data Catalog Network 48 Working with PARMON PARMON User Manual Figure 44 Kernel Data Catalog Memory The right half of the window lists the group and the nodes within that group Unlike other statistics cataloging does not support multiple nodes in a single window In case multiple nodes have to be cataloged each node should be separately selected and the desired parameter chosen In such cases the statistics too are projected on separate windows The top left corner of the 3 4 5 1 CPU Catalog The CPU catalog option catalogs the utilization statistics of the CPU over a period of 5 minutes with refresh rates of 2 3 4 or 5 seconds Figure 40 depicts a sample CPU catalog screen Working with PARMON 49 PARMON User Manual Figure 45 CPU Utilization Curve Syscalls 50 Working with PARMON PARMON User Manual Figure 46 CPU utilization curve Working with PARMON 5I PARMON User Manual Figure 47 CPU Utilization Context Switch The functioning of the Stop button is similar to that described in the section on CPU Usage The window can be closed by selecting the Close button 3 5 Miscellaneous Features Apart from the monitoring functions the PARMON tool can provide certain information like a node s hardware and software configuration The tool also allows managers users to execute general UNIX commands on the n
5. The Title Bar The contents of the title bar indicate the monitoring feature that is currently being accessed 3 1 2 The Icons The monitoring functions of the tool have been divided into the following categories e Aggregation e Message Viewer e Process Viewer e Kernel Data Catalog e Logical View Selecting the corresponding icon on the main screen can access each of these categories of functions Apart from the monitoring functions the tool provides for integration with SunVTS diagnostics tool from Sun Microsystems To invoke the SunVTS tool simply click on the icon named On line diagnostics 3 1 3 Menus All the functions accessible from icons are also accessible from the menus on the main screen Note however that setting up of the tool can be accomplished only from the Setup menu An interesting feature of the menu is that it can be detached torn off into a separate window A menu can be torn off by dragging pull with the left mouse button down it at the serration This feature allows for easier and faster access to menus that are frequently selected Remember that clicking on the tear off menu option is the same as clicking the respective menu from the main window and then selection the corresponding option Figure 2 depicts the appearance of the file tear off menu 10 Working with PARMON PARMON User Manual File Tear off Users Figure 2 File Tear off menu Each of the other menus too ca
6. is tedious and challenging task There is need of a tool which allows system administrator for effec tive monitoring of a huge system such as PARAM OpenFrame with ease PARMON is the answer or solution Unlike other monitoring systems PARMON is designed using latest Java Computing Technologies and offers portable flexible and comprehensive environment for monitoring of large clusters PARMON follows client server methodology and provides transparent access to all nodes to be monitored from a monitoring machine C DAC HPCC system software PARMON allows you to monitor activi ties and resource utilization of various components of clusters of workstations COW or Clusters of PCs COP running Solaris It monitors the machine both at node level and at the entire il Close Figure 56 Help More Help Figure 57 gives copyright information 60 Working with PARMON PARMON User Manual mE 8884 e O 888 amp 8 8 8 8888 88888 PERRE O 28 8 888 8 PARMON Team HPCC System Software Development Group Centre for Development of Advanced Computing 2 1 Ramanashree Plaza Brunton Road Bangalore Karnataka India PIN 560 025 Email parmon cdacb ernet in URL http cdach ernet in parmon Close Figure 57 Help About Working with PARMON 6l
7. pathname gt lt separator gt lt pathname gt where pathname is a path specification to the directory that java searches to locate the default libraries user compiled classes and archives Separator is a character that separates multiple path specifications Under the UNIX environment the separator is a colon and under the DOS Windows95 environment the separator is a semi colon Resource amp Setup Requirements 5 PARMON User Manual Eg Since Java is dependent on the native operating system please refer to your Operating System documentation for details regarding the setting up of environment variables However the following examples are shown to give an idea Under the C shell csh in UNIX setenv CLASSPATH SCLASSPATH SHOME parmon parmon jar Under the Windows95 DOS Environment SET LASSPATH C JAVA LIB C PARMON PARM ON JAR Q b PATH Alter this variable to include the path of the java run time The procedure for defining updating this variable also is Operating System dependent Under the C shell csh in UNIX setenv PATH SPATH opt java bin or set path Spath opt java bin Under the DOS Windows95 environment SET PATH PATH3 C JAVA BIN The setup_files directory contains certain configuration host amp group information It is recommended that these files be not edited as it may lead to improper functioning of the client In case the configuratio
8. time and the end time If required a search string can be specified 3 To accept the search criterion select the Accept button Selecting the Close button causes the search criterion to revert to default 4 Select the Start Search button Messages logged since the start time on the start date till the stop time on the stop date are displayed The search can be stopped in the middle by selecting the Stop Search button In case a search string is specified only those messages that have been logged 38 Working with PARMON PARMON User Manual during the specified period date and time and having the specified search string are displayed 00 00 00 20 05 00 Figure 32 Criteria Dialog box The messages shown on screen may be saved on to a file on the local disk Follow the procedure listed below to save the messages to a local file 1 Select the menu sequence File gt Save As A window similar to that in Figure 33 is displayed Working with PARMON 39 PARMON User Manual Figure 33 The SaveAs Window 2 Specify the required filename and select the Save button Since java uses the native windowing system the appearance functionality of the window may be entirely different on different operating systems 3 4 3 Processes The processes running on a specified group node s can be viewed using the Process Viewer To view the processes running on a specific group node s follow the pro
9. 5 Memory Usage Statistics as a Pie Chart 3 4 1 4 Network Usage The Network Usage Statistic presents the ratio between the total number of packets transmitted successfully and the number of packets in error for the default network interface on the specified node group This statistic is presented in the form of a chart either PIE or BAR and is refreshed periodically The refresh period could be one of 5 10 15 and 20 seconds The Network Usage chart will be similar to the ones Figure 26 Figure 27 shown below Working with PARMON 33 PARMON User Manual Ed Hetwork Usage Statistics Figure 26 Network Usage Statistics as a Pie Chart The meaning and behaviour of the components of the window is similar to those described in the section on CPU Usage refer section 3 4 1 1 34 Working with PARMON PARMON User Manual Et Hieternek Usage Statistics Loo Reteshing time is satio 1f secs Stop Chose Figure 27 Network Usage Statistics as a Bar Chart 3 4 1 5 Usage Matrix System managers of large cluster environments prefer to observe Statistics of the entire system at a glance and then proceed to enquire about the statistics of groups nodes etc Usage matrix provides exactly this functionality PARMON is capable of depicting the usage of a cluster system in three ways a Group wise b Node wise within a group and Working with PARMON 35 PARMON User Manual c Proc
10. ON for you System Monitoring a Functions You will never regret having it Send Cancel Figure 54 Broadcast message window Select the Cancel button to close the window 3 5 4 Online Help Online help gives a detailed description about using the tool Online help can be access under the Help menu on the main screen Figure 55and Figure 56 depict sample Help windows 58 Working with PARMON PARMON User Manual E PARMON Overview PaRMON Cluster Monitoring System C DAC HPCC system software PARMON allows you to monitor activi ties and resource utilization of various components of clusters of workstations COW or Clusters of PCs COP running Solaris It monitors the machine both at node level and at the entire system level exhibiting a single system image The system activi ties monitored by PARMON are the following Process Activities System Log Activities Kernel Activities Components Physical and Logical Views Device Control Events Generation and Analysis Figure 55 Help Overview Working with PARMON 59 PARMON User Manual fej PARMON More Help PARMON A Cluster Monitoring System Introduction Clusters of workstations have off late become a cost effective solution for high performance computing PARAM OpenFrame is a large cluster of high performance workstations interconnected using low latency high bandwidth communication networks Moni toring such huge systems
11. ar Chart OK Figure 12 The Chart Type window 3 3 4 Changing color and other preferences HH 3 4 Monitoring the cluster Several important statistics of the nodes in a cluster could be monitored using the tool The statistics may be monitored under the following categories e Aggregation e Messages e Processes e Logical View e Kernel Data Catalog 20 Working with PARMON PARMON User Manual 3 4 1 Aggregation This set of monitoring functions can be used to continuously monitor the usage of resources available on the node namely the CPU Memory Disk and Network This statistic can be viewed either in the form of a PIE chart or a BAR chart During a monitoring session it is also possible to change the type of chart that is projected To opt for Aggregation statistics select the icon named Aggregation in the main screen A pop up menu similar to the one depicted in the figure Figure 13 below is displayed PARMON A Cluster Monitoring System Figure 13 Popup menu under Aggregation To view Select on popup menu CPU Statistics CPU Usage Disk Statistics Disk Usage Memory Statistics Memory Usage Network Statistics Network Usage Usage Matrix Usage Matrix Selecting one of the pop up menu options displays a node selection window similar to that shown in Figure 14 Working with PARMON 21 PARMON User Manual Node Selection Window The Node Selection Windo
12. c node group This statistic is presented in the form of a chart either PIE or BAR and is refreshed periodically The refresh period could be one of standard 5 10 15 and 20 seconds or user defined The CPU Usage chart is similar to the ones Figure 16 Figure 17 shown below The Bar Figure 16 and Pie Figure 17 charts depict the usage graphically with the lower and higher extremes being 0 and 100 respectively The view can be switched from Pie to Bar and back dynamically To do so select the menu option Chart and opt for the desired chart type Refer also section 3 3 3 for defining the default chart type The refresh time interval defines the mean time between successive updates lower the refresh time faster the update The refresh time specified is in seconds The refresh interval can be redefined by selecting the TimeSet menu option and choosing the required refresh interval If desired the refresh cycle can be suspended and resumed The Stop button can be selected if one wishes to suspend the refresh When the refresh is suspended the Stop button is replaced by the Start button 24 Working with PARMON PARMON User Manual Figure 16 CPU Usage statistics as a Bar Chart The dialog window at the bottom of the window displays the status of the latest change refresh interval stop start etc affected on the window Working with PARMON 25 PARMON User Manual Figure 17 CPU Usage as a Pie Char
13. ce Setup Group gt Remove pops up the Group Remove window Figure 11 The Groups window on the left displays the list of defined groups Select the group definition s that needs to be deleted removed from the list The selected group definitions are shaded blue Selecting the Remove button will delete the definition Clicking on a selected group deselects it Selecting the Remove button deletes the selected groups Note that multiple groups can be selected and deleted at once 18 Working with PARMON PARMON User Manual To view the definition of a group select the required group in the Group Name list box The Selected Nodes window updates with the nodes belonging to that particular group To modify the definition of a group select the required group and select deselect the nodes in the Selected Nodes window Figure 11 The Group Remove window 3 3 3 Specifying Parameters At present the tool supports only a single parameter i e the Chart Type parameter Selecting the menu sequence Setup Parameters pops up the Chart type window Figure 12 where the preferred char type may be selected The preferred type Working with PARMON 19 PARMON User Manual may be either Pie Chart or Bar Chart may be selected To select a type click on the radio button beside the preferred chart type The preferred chart type is indicated by a filled black radio button Ei Chart Type Mm x ChartType B
14. ce for the server program e At least 32 MB of system RAM 64 MB recommended 2 1 2 Client Program A system can function as a client if it fulfils the following minimum requirements e Capable of running Java run time ver 1 1 4 or higher e The client system is connected to the network and TCP IP stack supporting sockets is configured on it e About 500 KB of free disk space for the client program Resource amp Setup Requirements 3 PARMON User Manual 2 2 Setting up the tool 2 2 1 Server Program The Server Program can be setup by copying the binary named parmond available in the distribution This program can be invoked so as to bind to a specified port The syntax for executing the server is parmond port id amp Where port id is an integer number that specifies a valid unused port In case the port number is not specified port 4000 is assumed by default The optional amp character forces the server to be run in the background Eg parmond 45 In case the port is in use by another application the server program exits immediately In such a case the server program has to be rerun so as to bind to an unused port It is not necessary to run several instances of the server program on a particular workstation in a cluster Multiple clients can access a running server program However it is necessary to bind the server to the same port on all the nodes in a cluster if they have to be
15. cedure listed below 1 Select the icon Message Viewer The Node Selection window similar to that depicted in Figure 14 is displayed 40 Working with PARMON PARMON User Manual 2 Select the required group node s and select the OK button A window similar to that shown in Figure 34 is displayed The processes are by default ordered on the process names It is also possible to order the processes on process number owner s username owner s user ID or size by selecting the Sort menu option and one of By Number By Username By UID or By Size respectively from the pop up menu It is possible to specify a refresh interval for updating the processes running on the specified group node s Specifying the refresh interval is similar to that defined in the section on CPU Usage Itis also possible to suspend and restart the refresh cycle The list of processes displayed on the window can be saved on to a file on the local disk by selecting the menu sequence File Save As Apart from the Save As and the Timeset menus whose behaviour is the same as that explained in the previous section refer section 3 4 1 1 the processes window provides for an additional Sort menu The Sort menu allows for sorting processes based on it names ID owner name owner ID and size memory image size Note It is recommended that the refresh be stopped before ordering the processes because refresh will cause the ordering to be reset to defa
16. d administrators managers several privileged UNIX commands are listed in the listbox one of which can be selected Issue Commands j m Figure 51 Issue Commands window In case a privileged command is issued a window similar to that depicted in Figure 52 is displayed To successfully execute such commands enter the root password and select the OK button RABEL Figure 52 Password window If the root password entered is not valid an error message is displayed within the same window and re entry of password is allowed The error message displayed is similar to that shown in Figure 53 56 Working with PARMON PARMON User Manual eB Message Box Root Password forganga Password is not correct Try again es ene Figure 53 Password Error message box Note In case a privileged command needs to be executed on multiple nodes the tool prompts for the root passwords of each of the nodes separately 3 5 3 Broadcasting Messages To broadcast messages select the procedure described below 1 Select the menu sequence File gt Broadcast Message 2 From the Node selection window select the group node s where the message should be broadcast A window similar to that depicted in Figure 54 is displayed 3 Enter the message and select the Send button Working with PARMON 57 PARMON User Manual E Message Broadcast Eg Enter Message to Broadcast Thank you for Patronizing PARM
17. e in percentage of the User Disk Area refer section 3 3 1 1 on the specific node group This statistic is presented in the form of a chart either PIE or BAR and is refreshed periodically The refresh period could be one of Working with PARMON 29 PARMON User Manual standard 5 10 15 and 20 seconds or user defined The Disk Usage chart will be similar to the ones Figure 22 Figure 23 shown below Figure 22 Disk Usage Statistics as a Bar Chart The meaning behaviour and functionality of the components of the window are similar to those described in the section on CPU Usage refer section 3 4 1 1 30 Working with PARMON PARMON User Manual Figure 23 Disk Usage Statistics as a Pie Chart 3 4 1 3 Memory Usage The Memory Usage Statistic presents the usage in percentage of the system memory RAM on the specific node group This statistic is presented in the form of a chart either PIE or BAR and is refreshed periodically The refresh period could be one of 5 10 15 and 20 seconds The Memory Usage chart will be similar to the ones Figure 24 Figure 25 shown below Working with PARMON 31 PARMON User Manual Figure 24 Memory Usage Statistics as a Bar Chart The meaning behaviour and functionality of the components of the window are similar to those described in the section on CPU Usage refer section 3 4 1 1 32 Working with PARMON PARMON User Manual Figure 2
18. essor wise within a node The group wise usage statistics window is similar to that depicted in Figure 28 Group Name Num Of Nodes CPU Usage Disk Usage Figure 28 Group wise Usage Matrix In the figure depicted above selecting a group shaded rose displays another matrix where the usages of the individual nodes in the selected group are displayed Figure 29 a a s a ee Figure 29 Node wise usage statistics of a group To view the usage statistics of the processors within a node select the CPU usage statistic shaded rose displayed adjacent to the required node The CPU usage statistics of the processors in a node appear similar to that depicted in Figure 30 36 Working with PARMON PARMON User Manual yamuna CPUS File TimeSet CPU Name Usage Cpu 1 14 __cpu2 11 Figure 30 Processor wise usage statistics of a node 3 4 2 Messages Messages critical to a system are usually displayed on the console These messages indicate the health of a system The messages displayed on the console are logged by the operating system to aid in diagnosing problems with the system System Managers frequently refer to these logs to diagnose the overall health of a system and take corrective action whenever necessary In large cluster systems it is an inhuman task to constantly keep a tab on the messages generated by the system The option to scan through the logs of an individual node s in a c
19. hat belong to that group and select the OK button Figure 7 The Group Create window Working with PARMON 15 PARMON User Manual Group Name is the symbolic name assigned to a group of nodes Similar to node names group names too cannot be empty strings Also group names cannot have embedded spaces and two group cannot have identical names If a group name is null or has embedded spaces an error message is displayed The error message is similar to that depicted in the following picture Figure 8 Group Name error message Null Names 16 Working with PARMON PARMON User Manual Figure 9 Group Name error message embedded blanks To select the nodes select the checkbox toggle towards the left of the node A selected node is indicates a tick mark in the checkbox To deselect a node simply click on a selected one Note There are two predefined groups namely Entire System and Individual Nodes These groups may be used to select all the nodes that have been defined or just a single node respectively 3 3 2 2 Group Modification To view and or modify the definition of a group select the menu sequence Setup gt Group gt View Modify The figure Figure 10 Working with PARMON 17 PARMON User Manual that follows depicts the appearance of the Group View Modify window Figure 10 Group View Modify window 3 3 2 3 Group Deletion Selecting the menu sequen
20. lowing window Figure 5 is displayed where information relating to the particular node can be provided The information required are 12 141 63 19F Figure 5 Node Create Delete window Node name the symbolic name by which a node is referred to Use of the actual hostname assigned to the node is strongly recommended though not compulsory Node Name cannot be an empty string null Also node names cannot have embedded spaces IP address the IP internet protocol address of the node User Disk Area partition of the disk that needs to be continuously monitored Refer to the section on aggregation Refer 3 4 1 for more information The user disk area can either be keyed in or selected from the dropdown listbox provided Select Add Node to add the definition of the node The window on the right Nodes Box shows the details of the nodes that have already been defined Working with PARMON 13 PARMON User Manual 3 3 1 2 Node Deletion It is also possible to delete nodes that have already been defined To delete a node s simply select the node s to be deleted in the Nodes Box and select Remove Node Whenever a node is selected that node is shaded blue in the Nodes Box Clicking on a selected node deselects it Note that it is possible to select multiple nodes to delete The following figure Figure 6 depicts several selected nodes Node Create Delete Window fa Node Name I Nodes Box rn
21. luster system can go a long way in easing the managerial task of such managers Selecting the icon Message Viewer on the main screen pops up the Node selection window To view the messages from a node or group select the appropriate group or nodes Note that even after selecting a group it is possible to select or deselect members of that group Follow the procedure defined in Creating Groups refer section 3 3 2 1 to select or deselect nodes Selecting the Individual Nodes group lists all the defined nodes of which at most one node could be selected Selecting the OK button displays a window similar to that depicted below Working with PARMON 37 PARMON User Manual Messages on Param Openframe ia File Search Criteria Start Search yamuna sendmail 10517 NOQUEUE yamuna sendmail 10866 NOQUEUE yamuna sendmail 10873 NOQUEUE edacb sendmail 4285 NOQUEUE edacb sendmail 6778 NOQUEUE edacb sendmail 7447 NOQUEUE Figure 31 The Message window The Messages window displays the messages logged in the default messages file var adm messages By default only messages logged since 0 00 hours on the current day are displayed In case other messages are required a search criterion has to be defined To define search criterion follow the procedure detailed below 1 Select the button Search Criteria A window similar to that depicted in Figure 32 is shown 2 Enter the start date end date start
22. n information is corrupt the files in this directory could be deleted and the tool could easily be configured again The client program can then be invoked by executing the following command java parmon port id 6 Resource amp Setup Requirements PARMON User Manual where port id is the port at which the client program talks with the server program Note that this port must be the same as that specified for the server Tip The environment variables can be initialized in the default boot up scripts or in the login scripts For more details regarding the environment variables please refer to your Operating System s documentation Resource amp Setup Requirements 7 PARMON User Manual Chapter 3 Working with PARMON The PARMON client program brings up a GUI when invoked The GUI comprises of menus and icons that may be selected to configure and monitor the cluster environment 3 1 Components of the GUI The main screen presented by the tool is similar to that depicted in the figure below Figure 1 PARMON A Cluster Monitoring System Title Bar File Monitor View Setup Users Processes Menu Broadcast Message Issue Commands E M duly W aa Aggregation Message Viewer Process Viewer Kernel Data Catalog ca cage y ka 2 Cai Icon Physical View Logical View Online Diagnostics Figure 1 The Main Screen Working with PARMON 9 PARMON User Manual 3 1 1
23. n be torn off in a similar manner Figure 3 depicts the appearance of the setup view and monitor tear off menus Figure 3 Main Screen with Tear off Menus 3 2 Setting up PARMON The GUI has been designed to aid comprehensive monitoring of the nodes of a cluster However Prior to employing the tool for monitoring purposes the client program must be configured to work in the target cluster environment Setting up PARMON involves defining the nodes groups of systems available in the cluster Follow the instructions detailed below to configure the client program Working with PARMON Il PARMON User Manual 3 3 Configuring the PARMON client 3 3 1 Node A cluster is defined as a group of independent workstations connected by means of a high speed network Each independent workstation is called a node The client program requires certain information about the nodes to be monitored like its name to ip address Internet Protocol address map user disk area etc Note that without this basic information the tool cannot be used for monitoring 3 3 1 1 Node Addition Node information can be defined by the following the steps detailed below On the main screen select click the left mouse button the Setup menu From the sub menu Figure 4 that is displayed select Node PARMON A Cluster Monitoring System Figure 4 Setup sub menu 12 Working with PARMON PARMON User Manual On selecting Node the fol
24. odes 3 5 1 1 The View Menu The View menu has options that enable a user manager to view the System Configuration and System Information The view menu has several sub menu options as depicted in Figure 48 52 Working with PARMON PARMON User Manual Selecting the menu sequence View SysConf and subsequently selecting a group node s displays the system configuration of the selected nodes Figure 49 depicts the sample output of the SysConf option Figure 48 View sub menu Working with PARMON 53 PARMON User Manual f4 System Configuration of PARAM OpenFrame nodes Figure 49 System Configuration window To view the System information select the menu sequence View SysInfo and subsequently choosing a group node s displays the System Information Figure 50 depicts a sample System Information window 54 Working with PARMON PARMON User Manual Figure 50 System Information window 3 5 2 Executing UNIX commands To execute standard UNIX commands on a group node s follow the procedure described below 1 Select the menu sequence File Issue Commands A node selection window pops up 2 Select the group node s on which the command is to be executed Select the OK button A window similar to that depicted in Figure 51 is displayed Working with PARMON 55 PARMON User Manual 3 Key in the command to be executed and press the OK button To ai
25. oduction 1 PARMON User Manual monitored The client program is based on the java and thus ensures compatibility across hardware and software platforms The tool allows a manager to monitor the nodes in a cluster from a single client doing away with the requirement of opening sessions on each of the nodes 1 2 Features of the Tool e Cross platform compatible client e User friendly GUI e Multi user support e Comprehensive monitoring support 1 3 Manual Organization The manual has been divided into several chapters e Introduction This chapter briefly introduces the cluster computing model and the need for cluster monitoring systems e Resource amp Setup Requirements This chapter details the software and hardware requirements and the procedure to setup the application e Working with PARMON This chapter details the components of the user interface along with essential details about configuration and use of PARMON 2 Introduction PARMON User Manual Chapter 2 Resource amp Setup Requirements The PARMON tool is based on the Client Server model in keeping with the concept of the Open architecture The server program is an ELF 32 bit MSB SPARC executable binary and the client program is a Java based client 2 1 Resource Requirements 2 1 1 Server Program The Server Program requires the following minimum requirements e A SuperSPARC UltraSPARC workstation e About 250 KB of free disk spa
26. t Figure 18 shows the Timeset sub menu and Figure 21 depicts the Chart sub menu From the sub menu it is possible to opt for a standard refresh time period or specify a user defined time interval 26 Working with PARMON PARMON User Manual CPU Usage Statistics Figure 18 CPU usage chart depicting Timeset sub menu Selecting the key in sub menu pops up a dialog window where the required refresh period can be specified The dialog window is similar to that depicted below Figure 19 The time input should be an integer value In case there is an error in input the message window similar to that depicted in Figure 20 is displayed Note Changes to refresh time take effect only after the following refresh i e the very next refresh after the time is redefined occurs at the period active currently Working with PARMON 27 PARMON User Manual Figure 19 Key in dialog window Figure 20 Invalid input error message window The chart sub menu allows the selection of the preferred type of chart Figure 21 depicts the appearance of the Chart sub menu 28 Working with PARMON PARMON User Manual CPU Usage Statistics Figure 21 CPU usage chart depicting Chart sub menu Selecting the Close button closes the window Note In case a node has multiple CPUs the average usage of all the CPUs will be computed and displayed 3 4 1 2 Disk Usage The Disk Usage Statistic presents the usag
27. ult Working with PARMON 41 PARMON User Manual Processes on PARAM OpenFrame Figure 34 Process Window depicting the Sort sub menu 3 4 4 Logical View The logical view of a system consists of a list of resources available on the system The resources have been classified under the following four heads CPU e Disk e Memory e Network To obtain the logical view of a specific group node s select the icon named Logical View on the main screen From the Node Selection window select a group or nodes of which the logical 42 Working with PARMON PARMON User Manual view is required and select the OK button A window similar to that shown in Figure 36 is displayed Figure 35 The Logical View window The following table lists the actions and the corresponding output Selecting Output CPU button Number of processors Name assigned to each processor Clock frequency Type of CPU for each processor Type of FPU of each processor Disk button File system information like Mount point Preferred Block size Fundamental Block Size Total number of blocks in the partition Total number of free blocks Free blocks available to an user Free blocks available to a Superuser Maximum length of a file name Maximum number of file nodes Type of file system The File system ID assigned to that file system Working with PARMON 43 PARMON User Manual Selecting
28. w allows the selection of one or more nodes groups from those already defined A group can be selected from the drop down list box Selecting a group selects all the nodes that belong to that group Further it is possible to deselect nodes Follow the procedure defined in Creating Groups refer 3 3 2 1 to select or deselect nodes Selecting the Individual Nodes group lists all the defined nodes of which at most one node could be selected Deselection is effective only so long as another group is not selected Also the group node s on which resource usage needs to be monitored should have the server running and bound to a valid port Select the group that needs to be monitored In case a group or multiple nodes are selected the statistics of all the selected nodes are projected within a single window Selecting the OK button displays the appropriate aggregation CPU Disk Network or Memory usage window 22 Working with PARMON PARMON User Manual Figure 14 The Node Group Selection Window Note In case the client program fails to communicate with PARMON server on the specified port on a given node workstation the client program displays the following message window Working with PARMON 23 PARMON User Manual Message Box Eg No Server Program is up Figure 15 Server Error Message window 3 4 1 1 CPU Usage The CPU Usage Statistic presents the usage in percentage of the CPU s on the specifi
29. yamuna 202 141 63 191 xport hor IP Address i ganga 202 141 63 190 eporthoms 3 192 xpo User Disk Area i fexporthome odach 202 141 ra 1 PTET 2 pesiar 202 144 63 110 tpat Add Node Remove Node EI Figure 6 The Node Create Delete window Selected Nodes Any number of such nodes can be defined Note however that each of the nodes should have a unique name and ip address 3 3 1 3 Node modification To change the definition of a node first delete the node and then create a new one by the same name 3 3 2 Group In a cluster environment there is a frequent need to select certain combinations of nodes for comparing statistics In order to simplify 14 Working with PARMON PARMON User Manual the selection of such combinations the group definition feature has been provided Using this feature it is possible to define symbolic group names to combinations of nodes Consequently selecting a group will enable the selection of all its constituent nodes of that group A single node may belong to several all groups 3 3 2 1 Group Creation Selecting the menu sequence Setup gt Group Create Figure 1 pops up the window depicted below Figure 7 where groups may be created Creating groups involves providing a symbolic name to a group and selecting the nodes that ought to be referred by that name In the screen depicted below simply provide a symbolic name and select the nodes t
Download Pdf Manuals
Related Search