- 28 Nov, 2017 3 commits
-
-
Marshall Garey authored
Bug 4406
-
Danny Auble authored
-
Marshall Garey authored
Bug 4405
-
- 27 Nov, 2017 17 commits
-
-
Tim Wickberg authored
Follow on to a37f8a1d for slurm.spec.
-
Pär Lindfors authored
Avoid duplicating this logic in the Source and setup definitions. Bug 4421.
-
Pär Lindfors authored
%systemd_preun_with_restart was a typo, replace with correct %systemd_postun_with_restart macro. Bug 4421.
-
Tim Wickberg authored
This reverts commit 29ac92fc. Restore the expected and long-standing practice of having the main slurm package own /etc/slurm. Bug 4421.
-
Tim Wickberg authored
-
Doug Jacobsen authored
Bug 4433.
-
Danny Auble authored
since the ABI could change from one version to the other. Bug 4435
-
Felip Moll authored
Bug 4431
-
Danny Auble authored
This primarily happens regularly if PrologFlags=contain Where the output has a nodename (and usually a number in it) Normal output is salloc: Granted job allocation 31159 1 but if the allocation is delayed you get output like salloc: Granted job allocation 31161 salloc: Waiting for resource configuration salloc: Nodes snowflake4 are ready for job 1 salloc: Relinquishing job allocation 31161 Where the test would key on the number in snowflake4 and fail. This change makes it so the number has to be the first thing seen.
-
Tim Wickberg authored
-
Pär Lindfors authored
With the change to dynamic linking for libslurmfull, all packages must match not just the API version but the exact release. Set this for all sub-packages in accordance with the Fedora packaging guidelines. Bug 4426.
-
Morris Jette authored
-
Dominik Bartkiewicz authored
bug 4356
-
Tim Wickberg authored
Continue to build with OpenSSL by default, which can be disabled with --without openssl. No behavioral change. Bug 4428.
-
Alejandro Sanchez authored
_top_priority() is used to determine at job allocation time (called from job_allocate()) if any other job has a higher priority than the specified job being allocated. If so, it sets the job state_reason to WAIT_PRIORITY. This patch skips the priority comparison if the two jobs being compared are pack job components from the same heterogeneous job. Bug 4400.
-
Morris Jette authored
Global replace operation made some undesired changes
-
Felip Moll authored
bug 4247
-
- 24 Nov, 2017 3 commits
-
-
Dominik Bartkiewicz authored
in _unpack_node_info_members(). Bug 4389
-
Danny Auble authored
Add special case for SELECT_PLUGIN_CRAY_CONS_RES, SELECT_PLUGIN_CRAY_LINEAR to select_get_plugin_id_pos() Bug 4389
-
Brian Christiansen authored
If a pending federated job exists on clusters 2 and 3 and squeue is run from cluster 1 then the active siblings can come and go depending on which cluster returns the job info first and depending if that cluster is the origin cluster or not. The origin cluster only knows where the active siblings are.
-
- 22 Nov, 2017 9 commits
-
-
Brian Christiansen authored
from status commands. Bug 4341
-
Danny Auble authored
Add in strong_alias calls so these functions will appear in libslurm and not just libslurmfull. Otherwise test7.3 can fail due to a missing symbol if a gpu gres is allocated to the job. Bug 4415.
-
Tim Wickberg authored
-
Felip Moll authored
Bug 4404.
-
Tim Wickberg authored
This setting causes /unix to be omitted from the xauth string. Simplify the regex to handle this by adding / to the earlier match and dropping the /unix pattern. Bug 4417.
-
Morris Jette authored
bug 4400
-
Dominik Bartkiewicz authored
bug 4256
-
Morris Jette authored
-
Dominik Bartkiewicz authored
Bug 4379
-
- 21 Nov, 2017 8 commits
-
-
Dominik Bartkiewicz authored
Can cause slurmstepd to crash, as rlimit_name was pointing to part of the free'd env_name variable. Bug 4409.
-
Artem Polyakov authored
This patch has fixed the problem to me. We are going to do some more verification later today and update. But I would appreciate if somebody else can test it as well. Signed-off-by:
Danny Auble <da@schedmd.com>
-
Morris Jette authored
There was a list of pending pack job records under consideration for scheduling by the backfill plugin that was not being cleared between interations of the backfill scheduler resulting in various scheduling anomalies. bug 4371, 4400
-
Morris Jette authored
An abort was triggered here due to a pack job start failure.
-
Danny Auble authored
code dealing with how we need to keep track of it. Bug 4405
-
Morris Jette authored
For heterogeneous job steps, the srun --open-mode option default value will be set to "append".
-
Morris Jette authored
Previous logic would fail if more than 2 pack groups implicitly specified in srun command.
-
Morris Jette authored
Previous logic would delay the initiation of pack jobs until all components were submitted. The new logic will defer pack job scheduling based upon a new "pack_job_offset" field in the job submit request and NOT set a begin_time in the future. This will eliminate the pack job scheduling reason value of "BEGIN_TIME". bugs 4369, 4400
-