fix: Backport slurmctld segfault patch

Signed-off-by: Brad Beam <[email protected]>

Author: bradbeam

images/patches/slurm/0024-slurmctld-segfault-epilog.patch

@@ -0,0 +1,12 @@
+diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
+index ec5cbfa81a..fc46c38956 100644
+--- a/src/slurmctld/job_mgr.c
++++ b/src/slurmctld/job_mgr.c
+@@ -16289,6 +16289,7 @@ extern bool job_epilog_complete(uint32_t job_id, char *node_name,
+ 	 * really started. Very rare obviously.
+ 	 */
+ 	if ((IS_JOB_PENDING(job_ptr) && (!IS_JOB_COMPLETING(job_ptr))) ||
++	    ((!job_ptr->node_bitmap_cg) && (!IS_JOB_COMPLETING(job_ptr))) ||
+ 	    (job_ptr->node_bitmap == NULL)) {
+ #ifndef HAVE_FRONT_END
+ 		uint32_t base_state = NODE_STATE_UNKNOWN;

images/patches/slurm/README.md

@@ -204,3 +204,8 @@ Notes from `pthead_detch` man
 In SLURM when a job fails due to not being able to meet the segment size requirements, the reason is `FAIL_BAD_CONSTRAINTS`. When a job is in this state, it is set to priority = 0, which is a held state. The scheduler will skip evaluating the job on future runs.
 
 This patch is to change it so that jobs that fail for unmet segment size requirements to not hold the job. So that if there are topology changes to the cluster, that can satisfy the job requirements, the job can still schedule. This will set the job reason to `Reason=Resources` instead of `Reason=BadConstraints`.
+
+### 0024-slurmctld-segfault-epilog.patch
+
+This patch backports the following commits to address an issue with slurmctld segfaults
+[SlurmCommit](https://github.com/SchedMD/slurm/commit/3d9f9647b416c47a238600e08b2887e4b8b93c2e)