- Apr 14, 2020
-
-
Brian Christiansen authored
Bug 8574 Signed-off-by:
Marshall Garey <marshall@schedmd.com>
-
Dominik Bartkiewicz authored
Reset pend_run_tasks in build_job_queue() see commit 5efab599 Bug 8731
-
Marshall Garey authored
build_job_queue() sets the array_task_id for the meta job of the job array when the job array gets split because of a burstbuffer or correspond dependency. This meant that for local dependencies the meta job would get tested as the final task of the array with an array_task_id that was not NO_VAL or INFINITY. However, we weren't re-sending out the dependency for the meta job to the remote cluster, so it wasn't getting the array_task_id. Bug 8574
-
Jason Bacon authored
Bug 8602.
-
- Apr 13, 2020
-
-
Albert Gil authored
Previously slurmd was discarding MPS if defined without File and the node had only 1 GPU. Now we link the MPS to that GPU by default. Previously slurmd was sending a MPS wihtout GPU RPC to the controller if that was set in conf files, and the controller was aborting due internal assertions. Now even if broken RPC is sent the controller won't abort but return Invalid Argument, and slurmds are no more sending broken RPCs but fatal with a clear message instead. Bug 8775 Co-authored-by:
Michael Hinton <hinton@schedmd.com>
-
Albert Gil authored
No functionality change. Bug 8775
-
Albert Gil authored
Bug 8775 Co-authored-by:
Michael Hinton <hinton@schedmd.com>
-
Marcin Stolarek authored
outside of the reservation allocation. Don't update exc_core_bitmap in job_test_resv for FLEX reservations If reservation is created with FLEX flag, then we shouldn't exclude any cores. Bug 8206.
-
Marcin Stolarek authored
-
- Apr 11, 2020
-
-
Ben Roberts authored
As well as the header in 'sacctmgr dump cluster'. Bug 8817.
-
Danny Auble authored
Allow a job that was once FAIL_[QOS|ACCOUNT] to be eligible again if the qos|account limitation is remedied. Bug 8134
-
Danny Auble authored
Currently no functional change. Bug 8134
-
Danny Auble authored
Bug 8134
-
Danny Auble authored
No functional change Bug 8134
-
Danny Auble authored
Clearing it makes it impossible to use if it becomes available in the future. Bug 8134
-
- Apr 10, 2020
-
-
Brian Christiansen authored
-
Colby Ashley authored
Bug 8675
-
Colby Ashley authored
Bug 8675
-
Ben Roberts authored
Bug 8829
-
Ben Roberts authored
Bug 8829
-
Jason Bacon authored
(Mistake introduced by Tim Wickberg on review.)
-
Jason Bacon authored
Everything is already exported under global, so a wildcard for local does nothing. (Except break the build for FreeBSD with LLVM 10.0.) Bug 8602.
-
- Apr 09, 2020
-
-
Nate Rini authored
Bug 8823.
-
Jason Bacon authored
Needed on FreeBSD. Bug 8602.
-
Jason Bacon authored
Needed on FreeBSD. Bug 8602.
-
Danny Auble authored
Signed-off-by:
Brian Christiansen <brian@schedmd.com>
-
Marcin Stolarek authored
The statement on the relation between tcp_syncookie and tcp_max_syn_backlog was wrong. Those were related years ago, but the behavior was opposite what was documented. Bug 8810.
-
Felip Moll authored
Calling acct_gather_profile_g_add_sample_data() during fini() causes a deadlock in acct_gather_profile_init(). Bug 8773
-
- Apr 08, 2020
-
-
Ben Roberts authored
Bug 8778
-
Nate Rini authored
-
Ben Roberts authored
Bug 7235
-
Marshall Garey authored
Trigger callbacks were initialized in open_slurmdbd_conn() after they were used in _open_slurmdbd_conn(). Instead we now initialize them inside of _open_slurmdbd_conn(). The consequence of the resume trigger callbacks (such as primary slurmdbd resumed operation) not happening on slurmctld restart was that the failure trigger callbacks (such as primary slurmdbd failure) would never get called until slurmdbd was restart while slurmctld was running. Bug 8198
-
Nate Rini authored
Bug 8787.
-
Nate Rini authored
Bug 8787.
-
Marshall Garey authored
No functional change. Bug 8492.
-
- Apr 07, 2020
-
-
Ben Roberts authored
Bug 8682.
-
Felip Moll authored
Bug 8804.
-
Nate Rini authored
Allows for non-SlurmUser/root job submission from slurmrestd. Bug 8787.
-
Nate Rini authored
Bug 8787.
-