[slurm-users] Slurm version 17.11.0 is now available

Danny Auble da at schedmd.com
Tue Nov 28 14:55:07 MST 2017


After 9 months of development and testing we are pleased to announce the 
availability of Slurm version 17.11.0!

As usual this can be downloaded from https://www.schedmd.com/downloads.php.

Thanks again for all the help and support to get this out the door. It 
was fun to see most of you at SC17!

Highlights of this release include

  -- Support for federated clusters to manage a single work-flow across 
a set of
     clusters.
  -- Support for heterogeneous job allocations (various processor types, 
memory
     sizes, etc. by job component). Support for heterogeneous job steps 
within a
     single MPI_COMM_WORLD is not yet supported for most configurations.
  -- X11 support is now fully integrated with the main Slurm code. 
Remove any
     X11 plugin configured in your plugstack.conf file to avoid errors being
     logged about conflicting options.
  -- Added new advanced reservation flag of "flex", which permits jobs 
requesting
     the reservation to begin prior to the reservation's start time and use
     resources inside or outside of the reservation. A typical use case 
is to
     prevent jobs not explicitly requesting the reservation from using those
     reserved resources rather than forcing jobs requesting the 
reservation to
     use those resources in the time frame reserved.
  -- The sprio command has been modified to report a job's priority 
information
     for every partition the job has been submitted to.
  -- Group ID lookup performed at job submit time to avoid lookup on all 
compute
     nodes. Enable with PrologFlags=SendGIDs configuration parameter.
  -- Add the configure option --with-shared-libslurm which will link to
     libslurm.so instead of libslurm.o thus reducing the footprint of 
all the
     binaries.
  -- In switch plugin, added plugin_id symbol to plugins and wrapped
     switch_jobinfo_t with dynamic_plugin_data_t in interface calls in
     order to pass switch information between clusters with different switch
     types.
  -- Changed default ProctrackType to cgroup.
  -- Changed default sched_min_interval from 0 to 2 microseconds.
  -- CRAY: --enable-native-cray is no longer an option and is on by default.
     If you want to run with ALPS please configure with 
--disable-native-cray.
  -- Added new 'scontrol write batch_script <jobid>' command to fetch a 
job's
     batch script. Removed the ability to see the script as part of the
     'scontrol -dd show job' command.
  -- Add new "billing" TRES which allows jobs to be limited based on the 
job's
     billable TRES calculated by the job's partition's TRESBillingWeights.
  -- Regular user use of "scontrol top" command is now disabled. Use the
     configuration parameter "SchedulerParameters=enable_user_top" to enable
     that functionality. The configuration parameter
     "SchedulerParameters=disable_user_top" will be silently ignored.
  -- Change default to let pending jobs run outside of reservation after
     reservation is gone to put jobs in held state. Added 
NO_HOLD_JOBS_AFTER_END
     reservation flag to use old default.




More information about the slurm-users mailing list