Table of Contents

Back to OAR versions

OAR 2.5.10

Release date: 2023-05-11


  • [oarnodesetting] Allow setting the suspected state
  • [oarstat] Allow to see other users' initial_request
  • [oarsub] ensure the inline job key is in a valid format, by adding \n
  • [oarsub] remove ssh inline keys from initial_request
  • [scheduler] add the SCHEDULER_RESOURCE_ORDER_ADV_RESERVATIONS_THRESHOLD option: advance reservations in a near future can use the same resources order as batch job, taking into account the standby and besteffort jobs)
  • [oarstat] Add a machine parseable output for -p option
  • [oarstat] Add a machine parseable output for -e (events) option
  • [oarstat] Exit with an error if -p and –sql options are used together
  • [oarnodes] Factorize SQL query used to get resources
  • [oarnodes] Factorize SQL query to get nodes' events
  • [oarnodes] Optimize OAR's resources querying (option -r)
  • [oarnodes] Optimize nodes' states querying
  • [oarnodes] Update usage and man page
  • [oarsub] add a reservation end time possibility to -r option
  • [oarsub] factorize the reservation parsing in interactive condition
  • [oarsub] be more flexible about the reservation dates parsing
  • [oarsub] add the 'now' keyword support to reservation request
  • [man] rework oarsub man page
  • [oarsub] update usage
  • [oarnodesetting] allow for unsetting the value of resource properties
  • [oarexec] add PASSIVE_JOB_HOOK_EXEC_FILE
  • [NodeChangeState] fix DB transaction error when resubmitting a job
  • [NodeChangeState] do not resubmit jobs for deploy/cosystem and if server_prologue error
  • [NodeChangeState] on prologue/epilogue error, only set the first node to suspected
  • [NodeChangeState] Do not suspect nodes for prologue/epilogue error of deploy/cosystem jobs
  • [oarsub] add the advance reservation validation hook feature
  • [oarnode] fix oarnode –sql if not resources to show
  • [Hulot] Add debug info regarding timeouts
  • [sarko] fix debug message for job frag
  • [accounting] change the behavior of accounting to be more efficient
  • [leon] optimize SQL query of get_to_kill_jobs subroutine
  • [MetaSched] Optimize querying of the last wake up for a list of node
  • [] Add constants to store often used OAR's job states
  • [Hulot] Fix unknown method oar_warning
  • [job_resource_manager] add support for AMD gpus
  • [oarsh] add support of AMD gpus
  • [oardodo] parametrize the OOM killer setup of oardodo
  • [NodeChangeState] factorize the resources state changing
  • [NodeChangeState] correctly lock and make transaction with PostgreSQL
  • [] factorize jobs fragging made by NodeChangeState
  • [] Add an lock_table_exclusive subroutine, for postgresql
  • [oarnodes] fix machine parseables output by returning an empty set
  • [oarnodes] fix uninitialized value with -r option
  • [oarnodes] fix issue when using non existing nodes in arguments
  • [oar.conf] annotate configuration lines with modules it is used for
  • [oarwalltime] always allow admins to perform walltime reduction
  • [oarapi] add a environment variable to configure the max cycle of cgi
  • [] do not specify a file name for stderr/stdout for interactive jobs
  • [judas] allow a custom instance name in notification mail subjects
  • [ssh] Remove deprecated option in sshd config
  • [core/scripts] add oar-node start/stop scripts to launch with systemd
  • [setup] Add systemd services unit for oar-node and oar-server
  • [Makefiles] Add support for systemd
  • [oar-database] Add option to skip the installation of the default admission rules when creating a new OAR database
  • [finaud/hulot] fix a race condition resulting in nodes being suspected when put in standby by hulot while finaud is running
  • [monika] show job events in job stats
  • [monika] add nodes filtering
  • [oarsub] add a -v switch (usable multiple times) to select verbosity
  • [misc] remove desktop_computing possibility in oarsub and API
  • [oarsub] add # before each outputed string (except the job ids) and better format of information and error messages
  • [criu checkpoint] added a timeout to exit when the resume fails
  • [oarapi] replace workdir by directory in OAR's rest API
  • [all] change the log format and make it configurable and various logs messages improvements
  • [oarstat] add jobs events to the gantt output (–gantt/-g)
  • [oarstat] make machine parseable output detailed view without using -f
  • [oarstat] add missing fields and uniq some when calling with –gantt
  • [oarstat] add a new output format '3' fixing several field names
  • [job_resource_manager] write env file without variable references
  • [oarwalltime] allow request cancellation
  • [oarwalltime] add –whole option
  • [oarwalltime] add a timeout for walltime change requests
  • [oarnodes/API] add functionality to provide the detailed jobs occupation on nodes
  • [api] better returned messages and add an OAR_IN_API env variable
  • [IO] make job creation with add_micheline_subjob() more consistent
  • [oarsub] enhance -C option to be more user friendly
  • [] optimize get_gantt_jobs_to_launch() SQL query
  • [oarapi] fix an error in logs due to redefined Dump() subroutine
  • [tools] add the pam_oar_adopt script

About cgroups

OAR does not use Linux CGroups v2. The operating system on nodes must enable CGroups v1 in order for OAR to function with its usual advanced features. You can enable the Linux CGroups v1 (if disabled by default in your distro) by adding to your kernel command line on all machines where the oar-node service runs:

systemd.unified_cgroup_hierarchy=false systemd.legacy_systemd_cgroup_controller=true

(see /boot/grub/grub.conf, or /etc/default/grub on all nodes) 

oar_2.5.10.txt · Last modified: 2023/11/22 11:08 by neyron
Recent changes RSS feed GNU Free Documentation License 1.3 Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki