- 15 Jul, 2015 7 commits
-
-
Morris Jette authored
Conflicts: src/common/slurm_step_layout.c src/plugins/select/cons_res/job_test.c src/plugins/select/cons_res/job_test.h src/plugins/select/cons_res/select_cons_res.c src/srun/libsrun/opt.c
-
Nathan Yee authored
-
Nathan Yee authored
-
Nathan Yee authored
Bug 1798
-
Morris Jette authored
If a job can only be started by preempting other jobs, the old logic could report the error: "cons_res: sync loop not progressing, holding job #" due to the usable CPUs and GRES needed by the pending job not matching. This change prevents the error message and job hold when job preemption logic is being used. The error message and job hold still take place for job scheduling outside of preemption, which will match CPUs and GRES at the beginning. bug 1750
-
Morris Jette authored
Under some conditions the select/cons_res plugin will hold a job, setting it's priority to zero and reason to HELD. The logic in slurmctld's main scheduling loop previously kept its priority at zero, but changed the reason from HELD to RESOURCES. This change leaves the proper job state as set by the select plugin. This may be related to bug 1750
-
Morris Jette authored
The backfill scheduler will periodically release locks for other actions. If a job is held during the time that locks were released, that job might still have been scheduled by the backfill scheduler (i.e. it failed to check for a job with a priority of zero). could be a root cause for bug 1750
-
- 14 Jul, 2015 10 commits
-
-
Danny Auble authored
-
Morris Jette authored
Conflicts: src/slurmctld/job_mgr.c
-
Morris Jette authored
Previous logic could fail to update some tasks of a job array for some fields. bug 1777
-
Danny Auble authored
-
Danny Auble authored
-
Morris Jette authored
the execve() call to launch an application can fail due to the system limit on open files or memory being reached. This adds retry logic. bug 1803
-
Morris Jette authored
Also add some cosmetic changes
-
Danny Auble authored
bind to the interface of return of gethostname instead of any address on the node which avoid RSIP issues in Cray systems. This is most likely useful in other systems as well.
-
Morris Jette authored
-
Morris Jette authored
Add level to switch table information logged by select plugin
-
- 13 Jul, 2015 4 commits
-
-
Morris Jette authored
-
Morris Jette authored
Old logic could purge a job record for a job that was in completing state (if there was also a lot of agent threads). This change prevents purging job records for completing jobs.
-
Morris Jette authored
Fix to job array update logic that can result in a task ID of 4294967294. To reproduce: $ sbatch --exclusive -a 1,3,5 tmp Submitted batch job 11825 $ scontrol update jobid=11825_[3,4,5] timelimit=3 $ squeue JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) 11825_3 debug tmp jette PD 0:00 1 (None) 11825_4 debug tmp jette PD 0:00 1 (None) 11825_5 debug tmp jette PD 0:00 1 (None) 11825 debug tmp jette PD 0:00 1 (Resources) A new job array entry was created for task ID 4 and the "master" job array record now has a task ID of 4294967294. The logic with the bug was using the wrong variable in a test. bug 1790
-
Gene Soudlenkov authored
Bug 1799
-
- 11 Jul, 2015 2 commits
-
-
Nathan Yee authored
Increase total backfill scheduler run time in stats_info_response_msg data structure from 32 to 64 bits in order to prevent overflow.
-
Nathan Yee authored
Replace logic of this form: if (x) { list_destroy(x); x = NULL; } with the simpler FREE_NULL_LIST(x) bug 1785
-
- 10 Jul, 2015 16 commits
-
-
Morris Jette authored
Conflicts: doc/html/download.shtml
-
Morris Jette authored
remove new capabilities added in comit ad9c2413 Leave the new logic only in version 15.08, which has related performance improvements in the slurmctld agent code, see commit 53534f49
-
Morris Jette authored
Modify slurmctld outgoing RPC logic to support more parallel tasks (up to 85 RPCs and 256 pthreads; the old logic supported up to 21 RPCs and 256 threads). This change can dramatically improve performance for RPCs operating on small node counts. bug 1786
-
Morris Jette authored
Change the case of "FAILURE" to "failure" in a test so we don't see upper case "FAILURE" on expected behaviour.
-
Danny Auble authored
aggregation.
-
Danny Auble authored
-
Danny Auble authored
little adding the extra hops, but if you are running a bunch of different small steps it can help out a lot on messages going to the controller.
-
Morris Jette authored
Correct "sdiag" backfill cycle time calculation if it yields locks. A microsecond value was being treated as a second value resulting in an overflow in the calcuation. bug 1788
-
Danny Auble authored
-
Danny Auble authored
-
Danny Auble authored
message that may come.
-
Danny Auble authored
-
Danny Auble authored
-
Danny Auble authored
-
Morris Jette authored
-
Morris Jette authored
-
- 09 Jul, 2015 1 commit
-
-
Morris Jette authored
The slurmctld logic throttles some RPCs so that only one of them can execute at a time in order to reduce contention for the job, partition and node locks (only one of the effected RPCs can execute at any time anyway and this lets other RPC types run). While an RPC is stuck in the throttle function, do not count that thread against the slurmctld thread limit. but 1794
-