OAR Documentation
Contents
1. So after a successful checkpoint if the job is resubmitted then all will go right and there will have no problem like file creation deletion Example oarsub checkpoint 600 signal 2 t idempotent path to prog So this job will send a signal SIGINT see man kill to know signal numbers 10 minutes before the walltime ends Then if everything goes well it will be resubmitted 6 6 How to submit a non disturbing job for other users You can use the besteffort job type Thus your job will be launched only if there is a hole and will be deleted if another job wants its resources Example oarsub t besteffort path to prog 7 OAR CHANGELOG 7 1 next version e Bugfix Fix a regression only for PostgreSQL in Drawgantt intro duced in the version 2 4 6 thanks to Yann Genevois 7 2 version 2 4 7 e Backport Debug checkpoint feature with cosystem or deploy jobs 7 3 version 2 4 6 e Fix the user variable used in oarsh When using oarsh from the frontal the variable OAR_USER was not defined in the environment and make oarsh unable to read the user private key file e Backport Bugfix 13434 reservation were not handled correctly with the energy saving feature e Draw Gantt Do not display Absent node in the future that are in the stanby sub state e Bugfix spelling error network_addess gt network_address 7 4 version 2 4 5 e backport node_change_state do not Suspect the first node of a job wh2. it must have exactly the same name that OAR has given in argument of the command JOBDEL_SOFTWALLTIME time after a normal frag that the system waits to retry to frag the job JOBDEL_WALLTIME time after a normal frag that the system waits before to delete the job arbitrary and suspects nodes LOG_FILE specify the path of OAR log file default var log oar log 30 Add wait in pingchecker to avoid zombies Better code modularization Remove node install part to launch jobs So it is easier to upgrade from one version to an other oarnodesetting must already be in stalled on each nodes if we want to use it Users can specify a method to be notified mail or script Add cpuset support Add prologue and epilogue script to be executed on the OAR server before and after launching a job Add dependancy support between jobs a option in oarsub In oarsub you can specify the launching directory d option In oarsub you can specify a job name n option In oarsub you can specify stdout and stderr file names User can resubmit a job option resubmit in oarsub It is possible to specify a read only database account and it will be used to evaluate SQL properties given by the user with the oarsub command more scecure Add possibility to order assigned resources with their properties by the scheduler So you can privilege some resources than others SCHED ULER_RESOURCE_
3. prop3 2 prop4 1 prop5 1 where names used in the hierarchical description prop first type of prop2 second type resources of resources add another resource hierarchy to the previous one sql SQL WHERE clause on the table of resources that filters resource sql2 SQL WHERE clause to apply on the second hierarchy request walltime 1 00 So we want to reserve 3 resources with the same value of the type prop2 and with the same property prop and these resources must fit sq To that possible resources we want to add 2 others which fit sq 2 and the hierarchy prop3 2 prop4 I prop5 1 Resource property hierarchy Tree example of a heterogeneous cluster SWITCH CORE 123 4 You can configure your own hierarchy with the property names that you want yaa This command reserves 2 cores on a cpu on a node oarsub d switch 2 nodes 1 cpu 1 core 2 on 2 different switchs so 2 computers oarsub I switch 1 This command reserves 1 switch entirely Figure 1 Example of a resource hierarchy and 2 different oarsub commands hierarchical_resources svg Examples oarsub 1 node 4 test sh the test sh script will be run on 4 entire nodes in the default queue with the default walltime oarsub stdout test12 jobid stdout stderr test12 jobid stderr nodes 4 test sh OAR_JOB_ID 702 same example than above but here the standard output of test sh will be written in
4. 10 connections could be too few 7 6 version 2 4 3 Hulot module now has customizable keepalive feature backport from 2 5 Added a hook to launch a healing command when nodes are sus pected backport from 2 5 Bugfix 9995 oaraccouting script doesn t freeze anymore when db is unreachable Bugfix 9990 prevent from inserting jobs with invalid username like an empty username Oarnodecheck improvements node is not checked if a job is already running New oaradmin option auto offset Feature request 10565 add the possibility to check the aliveness of the nodes of a job at the end of this one pingchecker 7 7 version 2 4 2 New Hulot module for intelligent and configurable energy saving Bug 9906 fix bad optimization in the gantt lib so bad scheduling 7 8 version 2 4 1 Bug 9038 Security flaw in oarsub notify option Bug 9601 Cosystem jobs are no more killed when a resource is set to Absent Fixed some packaging bugs API bug fixes in job submission parsing Added standby info into oarnodes s and available_upto info into resources uri of the API Bug Grid 5000 2687 Fix possible crashes of the scheduler Bug fix with MySQL DB Finaud suspected resources which are not of the default type Signed debian packages install oar keyring package 7 9 version 2 4 0 Fix bug in oarnodesetting command generated by oar_resources_init detect_resources Added a state option to oar
5. 4868 Add a parameter to specify what the nodes resource is a synomym for Network_address must be seen as an internal data and not used Scheduler add timeout for each job 1 4 of the remaining sched uler timeout 23 Bug 4866 now the whole node is Suspected instead of just the par where there is no job onto So it is possible to have a job on Suspected nodes e Add job walltime in seconds in parameter of prologue and epilogue on compute nodes oarnodes does not show system properties anymore e New feature container job type now allows to submit inner jobs for a scheduling within the container job e Monika refactoring and now in the oar packaging Added a table schema in the db with the field version reprensenting the version of the db schema Added a field DB_PORT in the oar config file Bug 5518 add right initialization of the job user name e Add new oaradmin command This will permit to create resources and manage admission rules more easily Bug 5692 change source code into a right Perl 5 10 syntax 7 15 version 2 2 12 Bug 5239 fix the bug if there are spaces into job name or project Fix the bug in lolib if DEAD_SWITCH_TIME gt 0 e Fix a bug in bipbip when calling the cpuset_manager to clean jobs in error Bug 5469 fix the bug with reservations and Dead resources Bug 5535 checks for reservations made at a same time was wrong e New feature local checks on nodes can be plugged in the
6. chart which shows job repartition on nodes in the time It is very useful to see cluster occupation in the past and to know when a job will be launched in the future 5 Mechanisms 5 1 How does an interactive oarsub work OARSUB _ oarsub I AAA bipbip AAA oarexec ae E oarexec init finished ob launche 7 give shell to user and set envi gt Ironnment Variables oarexe Cai ser sh coseno a M exit close shell END oar notif Oarexec END End of oarsub Ti me Enf of oarexec ecuser sh Trees Creer rr Figure 2 Interactive oarsub decomposition interactive_oarsub_scheme svg 5 2 Job launch For PASSIVE jobs the mechanism is similar to the INTERACTIVE one except for the shell launched from the frontal node The job is finished when the user command ends Then oarexec return its exit value what errors occured on the Almighty via the SERVER_PORT if DETACH_JOB_FROM_SERVER was set to 1 otherwise it returns directly 5 3 CPUSET The cpuset name is effectively created on each nodes and is composed as user_jobid OAR system steps 1 Before each job the Runner initialize the CPUSET see CPUSET definition with OPENSSH_CMD and an efficient launching tool Taktuk If itis not installed and configured TAKTUK_CMD then OAR uses an internal launching tool less optimized The processors assigned to this cpuset are taken from the defined database field by JOB_RESOURCE_
7. oarn odecheck mechanism Results can be asynchronously checked from the server taktuk ping checker e Add 2 new tables to keep track of the scheduling decisions gantt_jobs_predictions_log and gantt_jobs_resources_log This will help debugging scheduling troubles see SCHEDULER _LOG_DECISIONS in oar conf e Now reservations are scheduled only once at submission time Re sources allocated to a reservations are definitively set once the vali dated is done and won t change in next scheduler s pass e Fix DrawGantt to not display besteffort jobs in the future which is meaningless 7 16 version 2 2 11 e Fix Debian package dependency on a CGI web server e Fix little bug remove notification scheduled start time for Interac tive reservation e Fix bug in reservation take care of the SCHEDULER_JOB_SECURITY_TIME for reservations to check 24 Fix bug add a lock around the section which creates and feed the OAR cpuset Taktuk command line API has changed we need taktuk gt 3 6 Fix extra in the name of output files when using a job name Bug 4740 open the file in oarsub with user privileges S option Bug 4787 check if the remote socket is defined problem of timing with nmap Feature Request 4874 check system names when renaming prop erties DrawGantt can export charts to be reused to build a global multi OAR view e g DrawGridGantt Bug 4990 DrawGantt now uses the database localtime as its
8. toLaunch state and in Error Terminated Bug 7691 add array and array param file options parsing into the submitted script Fix also some parsing errors Bug 7962 enable resource property cm_availability to be manip ulated by the oarnodesetting command 7 12 version 2 3 3 Fix default admission rules case unsensitive check for properties used in oarsub Add new oaradmin subcommand oaradmin conf Useful to edit conf files and keep changes in a Subversion repository Kill correctly each taktuk command children in case of a timeout New feature array jobs option array on oarsub oarstat oardel oarhold and oarresume and file based parametric array jobs oarsub array param file in this version the DB scheme has changed If you want to upgrade your installation from a previous 2 3 release then you have to execute in your database one of these SQL script stop OAR before mysql DB mysql_structure_upgrade_2 3 1 2 3 3 sql postgres DB pg_structure_upgrade_2 3 1 2 3 3 sql 7 13 version 2 3 2 Change scheduler timeout implementation to schedule the maximum of jobs 22 Bug 5879 do not show initial_request in oarstat when it is not a job of the user who launched the oarstat command oar or root Add a event option to oarnodes and oarstat to display events recorded for a job or node Display reserved resources for a validated waiting reservation with a hint in their state Fix oarproperty prop
9. MANAGER_PROPERTY_DB_FIELD in the table resources 2 After each job OAR deletes all processes stored in the associ ated CPUSET Thus all nodes are clean after a OAR job If you don t want to use this feature you can but nothing will warranty that every user processes will be killed after the end of a job If you want you can implement your own cpuset management This is done by editing 3 files see also CPUSET installation e cpuset_manager pl this script creates the cpuset on each nodes and also delete it at the end of the job For more informations you have to look at this script there are several comments e oarsh OARSH this script is used to replace the standard ssh command It gets the cpuset name where it is running and transfer this information via ssh and the SendEnv option In this file you have to change the get_current_cpuset function e oarsh_shell OARSH_SHELL this script is the shell of the oar user on each nodes It gets environment variables and look at if there is a cpuset name So if there is one it assigns the current pro cess and its father to this cpusetname So all further user processes will remind in the cpuset In this file you just have to change the add_process_to_cpuset function 5 4 SSH connection key definition This function is performed by oarsub with the ssh_private_key and ssh_public_key options It enables the user to define a ssh key pair to connec
10. OAR Documentation User Guide Authors Capit Nicolas Emeras Joseph Address Laboratoire d Informatique de Grenoble Bat ENSIMAG an tenne de Montbonnot ZIRST 51 avenue Jean Kuntzmann 38330 MONTBONNOT SAINT MARTIN Contact nicolas capit imag fr joseph emeras imag fr Authors LIG laboratory Organization LIG laboratory Status Stable Copyright licenced under the GNU GENERAL PUBLIC LICENSE Dedication For users Abstract OAR is a resource manager or batch scheduler for large clus ters By it s functionnalities it s near of PBS LSF CCS and Condor It s suitable for productive plateforms and research experiments BE CAREFULL THIS DOCUMENTATION IS FOR OAR gt 2 3 0 PDF version OAR DOCUMENTATION USER pdf Table of Contents 1 OAR capabilities Description of the different commands 21 2 2 23 2 4 23 2 6 OIESLAE ie ost e E E E we A E e E a A Desktop computing Visualisation tools 4 1 4 2 Di ans oe sac sa ss he See Rope ee oe heed Bee hal Ge JR the Stee gered ae 8 G DAYA RG a eures nk a e a ERR ee eR Mechanisms 2 1 SA5 5 16 5 17 5 18 How does an interactive oarsub work o o Job lameli ox ee SO io ww a CPUGED 2c 00 whe bee ai bee ee Eee owe ed Sol connection Key definition gt o sece ss e men a ss io Suspendr s me oc ke Ae ea GAR a dey o sc ee eS So ee ae Checkpoint lt lt Ear FE RH aE REE ER Se Oe ES OCROTITE era Gigs Sec gts Sete ADAN ER Jebcepe
11. ORDER tag in oar conf file a command can be specified to switch off idle nodes SCHEDULER_NODE_MANAGER_SLEEP_CMD SCHEDULER_NODE_MANAGER_IDLE_TIME SCHEDULER_NODE_MANAGER_SLEEP_TIME in oar conf a command can be specified to switch on nodes in the Absent state according to the resource property cm_availability in the table re sources SCHEDULER_NODE_MANAGER_WAKE_UP_CMD in oar conf if a job goes in Error state and this is not its fault then OAR will resubmit this one 31
12. b you have to use t besteffort option of oarsub to specify that this is a besteffort job 15 Important a besteffort job cannot be a reservation If your job is of the type besteffort and idempotent oarsub t option and killed by the OAR scheduler then another job is automatically created and scheduled with same behaviours 5 16 Cosystem jobs This feature enables to reserve some resources without launching any program on cor responding nodes Thus nothing is done by OAR on computing nodes when a job is starting except on the COSYSTEM_HOSTNAME defined in the configuration file This is useful with an other launching system that will declare its time slot in OAR So yo can have two different batch scheduler When you submit a job you have to use t cosystem option of oarsub to specify that this is a cosystem job These jobs are stopped by the oardel command or when they reach their walltime or their command has finished They also use the node COSYSTEM_HOSTNAME to launch the specified program or shell 5 17 Deploy jobs This feature is useful when you want to enable the users to reinstall their reserved nodes So the OAR jobs will not log on the first computer of the reservation but on the DEPLOY_HOSTNAME So prologue and epilogue scripts are executed on DEPLOY_HOSTNAME and if the user wants to launch a script it is also executed on DEPLOY_HOSTNAME OAR does nothing on computing nodes because they normally will be
13. ckages e fixed performance issue on oarstat array_id index missing e fixed performance issue job_id index missing in event_log table e fixed a performance issue at job submission optimized a query and added an index on challenges table decisions 7 10 version 2 3 5 e Bug 8139 Drawgantt nil error Add condition to test the presence of nil value in resources table e Bug 8416 When a the automatic halt wakeup feature is enabled then there was a problem to determine idle nodes e Debug a mis initialization of the Gantt with running jobs in the metasched uler concurrency access to PG database 21 7 11 version 2 3 4 add the command oar_resources_init to easily detect and initialize the whole resources of a cluster oaradmin version now retrieve the most recent database schema number Fix rights on the schema table in postgresql Bug 7509 fix bug in add_micheline_subjob for array jobs job types Ctrl C was not working anymore in oarsub It seems that the signal handler does not handle the previous syntax SIG qdel Bug 7487 bad initialisation of the gnatt for the container jobs Fix bug in oarsh with the 1 option Bug 7634 test if the resource property value is effectively defined otherwise print a Bug 7674 remove the computation of the scheduler_priority field for besteffort jobs from the asynchronous OAR part Now the value is set when the jobs are turned into
14. e and delete before to the oaraccount ing command e cpuset are now stored in dev cpuset oar debian packaging configure and launch a specific sshd for the user oar e use a file descriptor to send the node list gt able to handle a very large amount of nodes every config files are now in etc oar e oardel can add a besteffort type to jobs and vis versa 28 7 34 version 2 0 2 add warnings and exit code to oarnodesetting when there is a bad node name or resource number change package version change default behaviour for the cpuset_manager pl more portable enable a user to use the same ssh key for several jobs at his own risk add node hostnames in oarstat f add accounting and u options in oarstat bug fix on index fields in the database syncro bug 2020 bug fix about server pro epilogue bug 2022 change the default output of oarstat Now it is usable bug 1875 remove keys in authorized_keys of oar on the nodes that do not correspond to an active cpuset clean after a reboot reread oar conf after each database connection tries add support for X11 forwarding in oarsub I and C debug mysql initialization script in debian package add a variable in oarsh for the default options of ssh to use more useful to change if the ssh version installed does not handle one of these options read oar conf in oarsh so admin can more easily change options in this script add support for X11 forwarding
15. e some free resources but the scheduler can decide to start your job later because they will have more free resources and the job walltime will be smaller 6 3 How can I submit a job with a non uniform description Example oarsub I 1 switch swl or switch sw5 switch 1 node 1 This example asks OAR to reserve all resources from the switch sw1 or the switch sw2 and a node on another switch You can see the syntax as a sub reservation directive 6 4 Can I perform a fix scheduled reservation and then launch sev eral jobs in it Yes You have to use the OAR scheduler timesharing feature To use it the reserva tion and your further jobs must be of the type timesharing only for you Example 1 Make your reservation oarsub r 2006 09 12 8 00 00 1 switch 1 t timesharing user This command asks all resources from one switch at the given date for the default walltime It also specifies that this job can be shared with himself and without a constraint on the job name 2 Once your reservation has begun then you can launch oarsub I 1 node 2 walltime 0 50 00 p switch nom_du_switch_sc t timesharing user So this job will be scheduled on nodes assigned from the pre vious reservation The timesharing oarsub command possibilities are enumerated in Timesharing 17 6 5 How can a checkpointable job be resubmitted automatically You have to specify that your job is idempotent
16. erty names are lowercase Fix OAR_JOB_PROPERTIES_FILE do not display system proper ties Add a new user command oarprint which allow to pretty print re source properties of a job Debug temporary job UID feature Add kill 9 on subprocesses that reached a timeout avoid Perl to wait something desktop computing feature is now available again ex oarsub t desktop_computing date Add versioning feature for admission rules with Subversion 7 14 version 2 3 1 Add new oarmonitor command This will permit to monitor OAR jobs on compute nodes Remove sudo dependency and replace it by the commands oardo and oardodo Add possibility to create a temporary user for each jobs on compute nodes So you can perform very strong restrictions for each job ex bandwidth restrictions with iptable memory management every thing that can be handled with a user id Debian packaging Run OAR specific sshd with root privileges un der heavy load kernel may be more responsive for root processes Remove ALLOWED_NETWORKS tag in oar conf added more com plexeity than resolving problems Ichange database scheme for the field exit_code in the table jobs Now oarstat exit_code line reflects the right exit code of the user passive job before even when the user script was not launched the exit_code was 0 which was BAD ladd DB field initial_request in the table jobs that stores the oarsub line of the user Feature Request
17. et is defined problem of timing with nmap Feature Request 4874 check system names when renaming prop erties e DrawGantt can export charts to be reused to build a global multi OAR view e g DrawGridGantt Bug 4990 DrawGantt now uses the database localtime as its time reference 7 22 version 2 2 10 e Job dependencies if the required jobs do not have an exit code O and in the state Terminated then the schedulers refuse to schedule this job e Add the possibility to disable the halt command on nodes with cm_availability value e Enhance oarsub S option more OAR parsed e Add the possibility to use oarsh without configuring the CPUSETs can be useful for users that don t want to configure there ssh keys 7 23 version 2 2 9 e Bug 4225 Dump only 1 data structure when using X or Y or D e Bug fix in Finishing sequence Suspect right nodes 7 24 version 2 2 8 e Bug 4159 remove unneeded Dump print from oarstat e Bug 4158 replace XML Simple module by XML Dumper one e Bug fix for reservation recalculate the right walltime e Print job dependencies in oarstat 7 25 version 2 2 7 e Bug 4106 fix oarsh and oarcp issue with some options erroneous leading space e Bug 4125 remove exit_code data when it is not relevant e Fix potential bug when changing asynchronously the state of the jobs into Terminated or Error 26 7 26 7 27 7 28 7 29 7 30 7 31 versio
18. he OAR database for the job and nodes will be Suspected by NodeChangeState If your job is check pointed and is of the type idempotent oarsub t option and its exit code is equal to 0 then another job is automatically created and scheduled with same behaviours 5 7 Checkpoint The checkpoint is just a signal sent to the program specified with the oarsub command If the user uses checkpoint option then Sarko will ask the OAR Perl script run ning on the first node to send the signal to the process SIGUSR2 or the one specified with signal You can also use oardel command to send the signal 5 8 Scheduling General steps used to schedule a job 1 All previous scheduled jobs are stored in a Gantt data structure 2 All resources that match property constraints of the job p option and indication in the from the 1 option of the oarsub are stored in a tree data structure according to the hier archy given with the 1 option 3 Then this tree is given to the Gantt library to find the first hole where the job can be launched 4 The scheduler stores its decision into the database in the gantt_jobs_predictions and gantt_jobs_resources tables See User section from the FAQ for more examples and features 5 9 Job dependencies A job dependency is a situation where a job needs the ending of another job to start OAR deals with job dependency problems by refusing to schedule depe
19. ically terminated if still running Walltime format is hour mn sec hour mn hour Ex nodes 4 cpu 1 walltime 2 00 00 S scanscript Batch mode only asks oarsub to scan the give script for OAR directives OAR 1 q queue lt queue gt Set the queue to submit the job to P property lt list gt Add constraints to properties for the job format is a WHERE clause from the SQL synta r reservation lt date gt Request a job start time reservation instead of a submission The date format is YYYY MM DD HH MM SS checkpoint lt delay gt Enable the checkpointing for the job A signe is sent DELAY seconds before the walltime on the first processus of the job signal lt sig gt Specify the signal to use when checkpointing Use signal numbers default is 12 SIGUSR2 ty type lt type gt Specify a specific type deploy besteffort cosystem checkpoint timesharing d directory lt dir gt Specify the directory where OAR will launch t command default is current directory project lt txt gt Specify a name of a project the job belongs t the job t be terminated to start n com use from a files instes ne instead of Warning the already exists y replace re the standart ly replace re the standart ly replace stead of Waitir ou must run e Waiting state n name lt txt gt Specif
20. ich was EXTERMINATED by Leon if the cpuset feature is con figured let do the job by the cpuset e backport OAREXEC ESRF detected that sometime oarexec think that he notified the Almighty with it exit code but nothing was seen on the server So try to resend the exit code until oarexec is killed 18 backport oar_Tools add in notify_almighty a check on the print and on the close of the socket connected to Almighty backport switch to bin bash as default some scripts currently need bash 7 5 version 2 4 4 Bug 10999 memory leak into Hulot when used with postgresql The leak has been minimized but it is still there DBD Pg bug Almighty cleans ipcs used by oar on exit Bugfix 10641 and 10999 Hulot is automatically and periodically restarted oar_resource_init bad awk delimiter There s a space and if the property is the first one then there is not a job suspend oardo does not exist anymore long long time ago Replace it with oardodo Bugfix oaradmin rules edition add was broken Bug 11599 missing pingchecker line into Leon Bug 10567 enabling to bypass window mechanism of hulot Back port from 2 5 Bug 10568 Wake up timeout changing with the number of nodes Backport from 2 5 oarsub when an admission rule died micheline returns an integer and not an array ref Now oarsub ends nicely Monika add a link on each jobid on the node display area sshd_config with nodes with a lot of core
21. installed on cluster login nodes So you must connect to one of these computers first 2 1 oarstat This command prints jobs in execution mode on the terminal Options 3 Job show informations only for the specified job eve f full show full informations s state show only the state of a job optimized query u user show informations for this user only g gantt show job informations between two date times e events show job events P properties show job properties accounting show accounting informations between two dates sql restricts display by applying the SQL where claus on the table jobs ex project pl1 D dumper print result in DUMPER format X xml print result in XML format Y yaml print result in YAML format backward compatible OAR 1 version like display V version print OAR version number h help show this help screen Examples oarstat oarstat j 42 f oarstat sql project pl oarstat s j 42 2 2 oarnodes This command prints informations about cluster resources state which jobs on which resources resource properties Options a shows all resources with their properties show only properties of a resource s shows only resource states 1 shows only resource list sql sql where Display resources which matches this sql where clause D formats outputs in Perl Dumper X for
22. ld for besteffort jobs from the asynchronous OAR part Now the value is set when the jobs are turned into toLaunch state and in Error Terminated Bug 7691 add array and array param file options parsing into the submitted script Fix also some parsing errors Bug 7962 enable resource property cm_availability to be manip ulated by the oarnodesetting command Added the standby information to a node state in oarnodes when it s state is Absent and cm_availability 0 Changed the name of cm_availability to available_upto which is more relevant e add a maintenance option to oarnodesetting that sets the state of a resource to Absent and its available_upto to 0 if maintenance is on and resets previous values if maintenance is off e added a signal option to oardel that allow a user to send a signal to one of his jobs e added a name field in the schema table that will refer to the OAR version name added a table containing scheduler name script and description Bug 8559 Almighty Moved OAREXEC_XXXX management code out of the queue for immediate action to prevent potential problems in case of scheduler timeouts e oarnodes oarstat and the REST API are no more making retry con nections to the database in case of failure but exit with an error in stead The retry behavior is left for daemons improved packaging try to install files in more standard places improved init script for Almighty into deb and rpm pa
23. le node is assigned then it kills all jobs from the user if all cpus are assigned e the same thing can be done with cpuset_manager_PAM pl as the script used to configure the cpuset More efficent if cpusets are con figured e debug cm_availability feature to switch on and off nodes automati cally depending on waiting jobs e reservations now take care of cm_availability field 7 33 version 2 1 0 e add oarcp command to help the users to copy files using oarsh e add sudo configuration to deal with bash Now oarsub and oarsh have the same behaviour as ssh the bash configuration files are loaded correctly e bug fix in drawgantt loose jobs after submission of a moldable one oar conf Thus admin can add some resources for each jobs like frontale node add possibility to use taktuk to check the aliveness of the nodes jobid is now replaced in stdout and stderr file names by the ef fective job id e change interface to shu down or wake up nodes automatically now the node list is read on STDIN add OARSUB_FORCE_JOB_KEY in oar conf It says to create a job ssh key by default for each job jobid is now replaced in the ssh job key name oarsub k e add NODE_FILE_DB_FIELD_DISTINCT_VALUES in oar conf that add SCHEDULER_RESOURCES_ALWAYS_ASSIGNED_TYPE into enables the admin to configure the generated containt of the OAR_NODE_FILE change ssh job key oarsub options behaviour e add options reinitializ
24. mats outputs in XML Y formats outputs in YAML Examples oarnodes oarnodes s oarnodes sql state Suspected 2 3 oarsub The user can submit a job with this command So what is a job in our context A job is defined by needed resources and a script program to run So the user must specify how many resources and what kind of them are needed by his application Thus OAR system will give him or not what he wants and will control the execution When ajob is launched OAR executes user program only on the first reservation node So this program can access some environment variables to know its environment SOAR_NODEFILE contains the name of a file which lists all reserved nodes for this job SOAR_JOB_ID contains the OAR job identificator SOAR_RESOURCE_PROPERTIES_FILE contains the name of a file which lists all resources and their properties SOAR_JOB_NAME name of the job given by the n option SOAR_PROJECT_NAME job project name Options I interactive Request an interactive job Open a login shel on the first node of the reservation instead running a script Es connect lt job id gt Connect to a running job As resource lt list gt Set the requested resources for the job The different parameters are resource propert registered in OAR database and walltime w specifies the duration before the job must be automat
25. n 2 2 6 e Bug fix job types was not sent to cpuset manager script anymore border effect from bug 4069 resolution version 2 2 5 e Bug fix remove user command when oar execute the epilogue script on the nodes Clean debug and mail messages format Remove bad oarsub syntax from oarsub doc e Debug xauth path e bug 3995 set project correctly when resubmitting a job e debug bash c on Fedora e bug 4069 reservations with CPUSET_ERROR remove bad hosts and continue with a right integrity in the database bug 4044 fix free resources query for reservation get the nearest hole from the beginning of the reservation e bug 4013 now Dead Suspected and Absent resources have different colors in drawgantt with a popup on them version 2 2 4 e Redirect third party commands into oar log easier to debug e Add user info into drawgantt interface e Some bug fixes version 2 2 3 e Debug prologue and epilogue when oarexec receives a signal version 2 2 2 e Switch nice value of the user processes into 0 in oarsh_shell in case of sshd was launched with a different priority e debug taktuk zombies in pingchecker and oar_Tools version 2 2 1 e install the allow_clasic_ssh feature by default e debug DB installer 27 7 32 version 2 2 e oar_server_proepilogue pl can be used for server prologue and epi logue to authorize users to access to nodes that are completely al located by OAR If the who
26. ndant jobs if their required job is in Terminated state and have an exit code 0 an error occured If the required job is resubmited its jobId is no longer the same and OAR updates the database and sets the job_id_required field to this new jobId for the dependant job 5 10 User notification This section explains how the notify oarsub option is handled by OAR e The user wants to receive an email The syntax is mail name Odomain com Mail section in the Configuration file must be present otherwise the mail cannot be sent The subject of the mail is of the form OAR TAG job_id Gob_name on OAR_server_hostname e The user wants to launch a script The syntax is exec path to script args OAR server will connect using OPENSSH_CMD on the node where the oarsub command was invoked and then launches the script with the following arguments job_id job_name TAG comments TAG can be e RUNNING when the job is launched e END when the job is finished normally e ERROR when the job is finished abnormally e INFO used when oardel is called on the job e SUSPENDED when the job is suspended e RESUMING when the job is resumed 5 11 Accounting aggregator In the Configuration file you can set the ACCOUNTING_WINDOW parameter Thus the command oaraccounting will split the time with this amount and feed the table accounting So this is very easily and faster to get usage statistics of the cluster We can see tha
27. ndencles ek sh ow RE RRR EERE Re RE ew RiserimKimcalion io bd wh hed a on Accoun AVETEPAIOF ans ar E ERE eR Ee Dynamic nodes coupling features o o Tines hanne oo kk a a EG we Bee BEY Container JODE oe ee 5445 bee ee ee Pea A ee B steffort jobs s ccor bas Ee ee e ad e COsystEm JODE oro coe E eai e e e A a Deploy WOR cio panas a GY die Ae aha k a eS Desktop computing sspe aras we 8 6 FAQ USER 7 6 1 6 2 6 3 6 4 6 5 6 6 Release policy oir a A E a ee How can I submit a moldable job How can I submit a job with a non uniform description Can I perform a fix scheduled reservation and then launch several jobs How can a checkpointable job be resubmitted automatically How to submit a non disturbing job for other users OAR CHANGELOG 7 1 T2 Fa 7 4 73 7 6 Next FOSO a a be AA dd venion 2AT o o ee a A OR GG a o VEO LAO ir ke A ha ee bem a A PIO ARE a Ae ke A E a Se A BD OR a Ge BOLE es Coon RWW Ww 15 15 15 15 15 16 16 16 16 16 16 17 17 17 A ZAZ oo a aa RG eS SS 18 Ea VEA LA N e a a ee eA ee BE eA hee Se Be 18 79 version 24 06 ce a OR O 18 EA AS bl i ee AG E Se CRAMER See 20 PAL Merson 238s ci Sew hw Hale Re E Sw amp Ae Be Se N 20 TAZ VERONZ o it e Boeke Bad sh ee 20 PAD VERIO 232 ca eRe EAE RE Oe eee eA 21 PAG Version ESE ci we Ok Addo wee SE A eee BE 21 FAD yesin ZAS coria a a HRA we Se a GOERS ER 22 ELO VERO BZ e
28. o ob a a a nd 23 Flt Version CI ok ko eh ee EEA Ee eS 23 T18 VERSION LL kick es ke ee a A BO ole 23 FAO YONA e A he RAE ORE EE ee 23 TL VERSION LE a So del OR A a GS ac LS 24 FED AENA A UE aa RE Ae eee eed A EES SE LS 24 Tal VESI ZATO 2 on ke ie hee ARE A 24 Fes VERNON 2 sf ks ae EA eA eee ee 24 FLA VERI ZAG rr wR we we BR Ole BER SS OR 24 Tar VERNON Zone a hw RE ARS ERR ee ewe A 25 FLO VERON 2258 624 55 8 Pah bee ee Dawe babe ee 25 Tied VERSION LA oes hat Se ho dt BG AS we OE Bk he Gee See e Aes 25 FEO VERSION aA co A he RAE ORE ee A 25 T29 VEIA 22 35 a A BR RS ac le 25 E a ARAN 25 Tal VEO Al a rito A ae a Oe it ey 26 Ll MORMON a A ee A a e ed 26 Pas vesion Ze 6c 4 ed is BL ee a BLE ee HM 26 POR VERSION AMS oes pa ee hs eR AR ERE Be aoe bg we 27 Foo wero Ze co ha ee A EH EE ES 27 1 OAR capabilities Oar is an opensource batch scheduler which provides a simple and flexible exploitation of a cluster It manages resources of clusters as a traditional batch scheduler as PBS Torque LSF SGE In other words it doesn t execute your job on the resources but manages them reservation acces granting in order to allow you to connect these resources and use them Its design is based on high level tools e relational database engine MySQL or PostgreSQL e scripting language Perl e confinement system mechanism cpuset e scalable exploiting tool Taktuk It is flexible enough to be suitable for production clusters and re
29. rebooted to install a new system image This feature is strongly used in the Grid5000 project with Kadeploy tools When you submit a job you have to use t deploy option of oarsub to specify that this is a deploy job 5 18 Desktop computing If you cannot contact the computers via SSH you can install the desktop computing OAR mode This kind of installation is based on two programs e oar cgi this is a web CGI used by the nodes to communicate with the OAR server via a HTTP server on the OAR server node e oar agent pl This program asks periodically the server web CGI to know what it has to do This method replaces the SSH command Computers which want to register them into OAR just has to be able to contact OAR HTTP server In this situation we don t have a NFS file system to share the same directories over all nodes so we have to use a stagein stageout solution In this case you can use the oarsub option stagein to migrate your data 6 FAQ USER 6 1 Release policy Since the version 2 2 release numbers are divided into 3 parts e The first represents the design and the implementation used e The second represents a set of OAR functionalities e The third is incremented after bug fixes 6 2 How can I submit a moldable job You just have to use several 1 oarsub option one for each moldable description By default the OAR scheduler will launch the moldable job which will end first So you can se
30. s that the job can be shared with all users but only one with the same name 4 timesharing userjob_name This indicates that the job can be shared only with the same user and one with the same job name See User section from the FAQ for more examples and features 5 14 Container jobs With this functionality it is possible to execute jobs within another one So it is like a sub scheduling mechanism First a job of the type container must be submitted for example oarsub I t container 1 nodes 10 walltime 2 10 00 OAR_JOB_ID 42 Then it is possible to use the inner type to schedule the new jobs within the previ ously created container job oarsub I t inner 42 1 nodes 7 oarsub I t inner 42 1 nodes 1 oarsub I t inner 42 1 nodes 10 Notes e In the case oarsub I t inner 42 1 nodes 11 This job will never be scheduled because the container job 42 re served only 10 nodes e t container is handled by every kind of jobs passive interactive and reservations But t inner cannot be used with a reserva tion 5 15 Besteffort jobs Besteffort jobs are scheduled in the besteffort queue Their particularity is that they are deleted if another not besteffort job wants resources where they are running For example you can use this feature to maximize the use of your cluster with multiparametric jobs This what it is done by the CIGRI project When you submit a jo
31. search experiments It currently manages over than 5000 nodes and has executed more than 5 million jobs OAR advantages No specific daemon on nodes No dependence on specific computing libraries like MPI We support all sort of parallel user applications Upgrades are made on the servers nothing to do on computing nodes CPUSET 2 6 linux kernel integration which restricts the jobs on assigned resources also useful to clean completely a job even parallel jobs All administration tasks are performed with the taktuk command a large scale remote execution deployment http taktuk gforge inria fr Hierarchical resource requests handle heterogeneous clusters Gantt scheduling so you can visualize the internal scheduler decisions Full or partial time sharing Checkpoint resubmit Licences servers management support Best effort jobs if another job wants the same resources then it is deleted automatically useful to execute programs like SETIO home Environment deployment support Kadeploy http kadeploy imag fr Other more common features Batch and Interactive jobs Admission rules Walltime Multi schedulers support Multi queues with priority Backfilling First Fit Scheduler Reservation Support of moldable tasks Check compute nodes Epilogue Prologue scripts Support of dynamic nodes Logging Accounting Suspend resume jobs 2 Description of the different commands All user commands are
32. stat to only get the status of specified jobs optimized query to allow scripting Added a REST API for OAR and OARGRID Added JSON support into oarnodes oarstat and oarsub New Makefile adapted to build packages as non root user add the command oar_resources_init to easily detect and initialize the whole resources of a cluster oaradmin version now retrieve the most recent database schema number Fix rights on the schema table in postgresql Bug 7509 fix bug in add_micheline_subjob for array jobs job types Ctrl C was not working anymore in oarsub It seems that the signal handler does not handle the previous syntax SIG qdel Fix bug in oarsh with the 1 option Bug 7487 bad initialisation of the gnatt for the container jobs Scheduler move the delete_unnecessary_subtrees directly into find_first_hole Thus this is possible to query a job like oarsub 1 1 nodes 1 core 1 nodes 4 core 2 no hard separation between each group 20 For the same behaviour as before you can query oarsub I l1 prop 1 nodes 1 core 1 f prop 2 no Bug 7634 test if the resource property value is effectively defined otherwise print a Optional script to take into account cpu core topology of the nodes at boot time to activate inside oarnodesetting_ssh Bug 7174 Cleaned default PATH from into oardodo Bug 7674 remove the computation of the scheduler_priority fie
33. t like a data warehousing information extraction method 5 12 Dynamic nodes coupling features We are working with the Icatis company on clusters composed by Intranet computers These nodes can be switch in computing mode only at specific times So we have implemented a functionality that can request to power on some hardware if they can be in the cluster We are using the field available_upto from the table resources to know when a node will be inaccessible in the cluster mode easily settable with oarnodesetting command So when the OAR scheduler wants some potential available computers to launch the jobs then it executes the command SCHEDULER_NODE_MANAGER_WAKE_UP_CMD Moreover if a node didn t execute a job for SCHEDULER_NODE_MANAGER_IDLE_TIME seconds and no job is scheduled on it before SCHEDULER_NODE_MANAGER_SLEEP_TIME seconds then OAR will launch the command SCHEDULER_NODE_MANAGER_SLEEP_CMD 5 13 Timesharing It is possible to share the slot time of a job with other ones To perform this feature you have to specify the type timesharing when you use oarsub You have 4 different ways to share your slot 14 1 timesharing This is the default behavior if nothing but timesharing is specified It indicates that the job can be shared with all users and every job names 2 timesharing user This indicates that the job can be shared only with the same user and every job names 3 timesharing job_name This indicate
34. t on their nodes So oarsh can be used on nodes of different clusters to connect each others if the same ssh keys are used with each oarsub So a grid reservation r option of oarsub on each OAR batch scheduler of each wanted clusters can be done with this functionality Example 39 ssh keygen f oar_key oarsub ssh_private_key cat oar_key ssh_public_key S cat oar_key f 5 5 Suspend resume Jobs can be suspended with the command oarhold send a SIGSTOP on every pro cesses on every nodes to allow other jobs to be executed Suspended jobs can be resumed with the command oarresume send a SIGSTOP on every suspended processes on every nodes They will pass into Running when assigned resources will be free IMPORTANT This feature is available only if CPUSET is configured You can specify 2 scripts if you have to perform any actions just after JUST_AFTER_SUSPEND_EXEC_FILE suspend and just before resume JUST_BEFORE_RESUME_EXEC_FILE Moreover you can perform other actions than send signals to processes if you want just edit the suspend_resume_manager pl file 5 6 Job deletion Leon tries to connect to OAR Perl script running on the first job node find it thanks to the file tmp oar pid_of_oarexec_for_jobId_id and sends a SIGTERM signal Then the script catch it and normally end the job kill processes that it has launched If this method didn t succeed then Leon will flush t
35. the file test12 702 stdout and the standard error in test12 702 stderr oarsub q default 1 node 10 cpu 3 walltime 2 15 00 p switch sw1 home users toto prog the home users toto prog script will be run on 10 nodes with 3 cpus so a total of 30 cpus in the default queue with a walltime of 2 15 00 Moreover p option restricts resources only on the switch sw1 oarsub r 2009 04 27 11 00 00 1 node 12 cpu 2 a reservation will begin at 2009 04 27 11 00 00 on 12 nodes with 2 cpus on each one oarsub C 42 connects to the job 42 on the first node and set all OAR environment variables oarsub p not host like nodename To exclude a node from the request oarsub I gives a shell on a resource 2 4 oardel This command is used to delete or checkpoint job s They are designed by their iden tifier Option sql delete checkpoint jobs which respond to the SQL where clause on the table jobs ex project pl c job_id send checkpoint signal to the job signal was definedwith signal option in oarsub Examples oardel 14 42 delete jobs 14 and 42 oardel c 42 send checkpoint signal to the job 42 2 5 oarhold This command is used to remove a job from the scheduling queue if it is in the Wait ing state Moreover if its state is Running oarhold can suspend the execution and enable other jobs to use its resources In that
36. time reference 7 17 version 2 2 10 7 18 7 19 7 20 7 21 Job dependencies if the required jobs do not have an exit code 0 and in the state Terminated then the schedulers refuse to schedule this job Add the possibility to disable the halt command on nodes with cm_availability value Enhance oarsub S option more OAR parsed Add the possibility to use oarsh without configuring the CPUSETs can be useful for users that don t want to configure there ssh keys version 2 2 9 Bug 4225 Dump only data structure when using X or Y or D Bug fix in Finishing sequence Suspect right nodes version 2 2 8 Bug 4159 remove unneeded Dump print from oarstat Bug 4158 replace XML Simple module by XML Dumper one Bug fix for reservation recalculate the right walltime Print job dependencies in oarstat version 2 2 7 version 2 2 11 Fix Debian package dependency on a CGI web server Fix little bug remove notification scheduled start time for Interac tive reservation Fix bug in reservation take care of the SCHEDULER_JOB_SECURITY_TIME for reservations to check 25 e Fix bug add a lock around the section which creates and feed the OAR cpuset e Taktuk command line API has changed we need taktuk gt 3 6 e Fix extra in the name of output files when using a job name Bug 4740 open the file in oarsub with user privileges S option e Bug 4787 check if the remote sock
37. ub 1 switch 1 cpu 5 which submit a job on 3 processors on the same switch Add a signal handler in oarexec and propagate this signal to the user process Support FOAR p options in user script Add in oar conf DB_BASE_PASSWD_RO for security issues it is possi ble to execute request with parts specified by users with a read only account like p option OARSUB_DEFAULT_RESOURCES when nothing is spec ified with the oarsub command then OAR takes this default resource description OAREXEC_DEBUG_MODE turn on or off debug mode in oarexec create tmp oar oar log on nodes FINAUD_FREQUENCY indicates the frequency when OAR launchs Finaud search dead nodes SCHEDULER_TIMEOUT indicates to the scheduler the amount of time after what it must end itself SCHEDULER_JOB_SECURITY_TIME time between each job DEAD_SWITCH_TIME after this time Absent and Sus pected resources are turned on the Dead state PROLOGUE_EPILOGUE_TIMEOUT the possibility to specify a different timeout for prologue and epilogue PRO LOGUE_EPILOGUE_TIMEOUT PROLOGUE_EXEC_FILE you can specify the path of the prologue script executed on nodes EPILOGUE_EXEC_FILE you can specify the path of the epilogue script executed on nodes GENERIC_COMMAND a specific script may be used in stead of ping to check aliveness of nodes The script must return bad nodes on STDERR 1 line for a bad node and
38. via oarsh change variable for oarsh OARSH_JOB_ID gt OAR_JOB_ID 7 35 version 2 0 0 Now with the ability to declare any type of resources like licences VLAN IP range computing resources must have the type default and a network_address not null Possibility to declare associated resources like licences IP ranges and to reserve them like others Now you can connect to your jobs not only for reservations Add cosystem job type execute and do nothing for these jobs New scheduler oar_sched_gantt_with_timesharing You can spec ify jobs with the type timesharing that indicates that this scheduler can launch more than job on a resource at a time It is possible to restrict this feature with words user and name For example t timesharing user name indicates that only a job from the same user with the same name can be launched in the same time than it Add PostGresSQL support So there is a choice to make between MySQL and PostgresSQL 29 New approach for the scheduling administrators have to insert into the databases descriptions about resources and not nodes Resources have a network address physical node and properties For example 1f you have dual processor then you can create 2 different resources with the same natwork address but with 2 different processor names The scheduler can now handle resource properties in a hierarchical manner Thus for example you can do oars
39. way a SIGINT signal is sent to every processes Options sql hold jobs which respond to the SQL where clause on the table jobs ex project pl1 Manage not only Waiting jobs but also Running one can suspend the job 2 6 oarresume This command resumes jobs in the states Hold or Suspended Option sql resume jobs which respond to the SQL where clause on the table jobs ex project pl1 3 Desktop computing If you want to compute jobs on nodes without SSH connections then this feature is for you On the nodes you have to run oar agent pl This script polls the OAR server via a CGI HTTP script Usage examples e if you want to run a program that you know is installed on nodes oarsub t desktop_computing path to program Then path to program is run and the files created in the oar agent pl run ning directory is retrieved where oarsub was launched e if you want to copy a working environment and then launch the program oarsub t desktop_computing s script sh The content of is transfred to the node script sh is run and every thing will go back 4 Visualisation tools 4 1 Monika This is a web cgi normally installed on the cluster frontal This tool connects to the DB gets relevant information then format data in a html page Thus you can have a global view of cluster state and where your jobs are running 4 2 DrawOARGantt This is also a web cgi It creates a Gantt
40. y an arbitrary name for a anterior lt job id gt Anterior job that mus this new one notify lt txt gt Specify a notification method mail or command to execute Ex notify mail name domai notify exec path to script args resubmit lt job id gt Resubmit the given job as a new one k use Job key Activate the job key mechanism i import job key from file lt file gt Import the job key to of generating a new one import Jjob key inline lt txt gt Import the job key to use inli generating a new one xport Job key to file lt file gt Export the job key to a file file will be overwritten if it the Sjobid pattern is automatical 0 stdout lt file gt Specify the file that will sto output stream of the job the jobid pattern is automatical E stderr lt file gt Specify the file that will sto error stream of the job the Sjobid pattern is automatical hold Set the job state into Hold in so that it is not scheduled y oarresume to turn it into th D dumper Print result in DUMPER format X xml Print result in XML format Y yaml Print result in YAML format h help Print this help message V version Print OAR version number Wanted resources have to be described in a hierarchical manner using the I syn tax option Moreover it is possible to give a specification that must be matched on properties So the long and complete syntax is of the form sqll prop1 1 prop2 3 1 sql2
Download Pdf Manuals
Related Search