# Conflicts:
#	src/common/gres.c

to prevent underflow.

Bug 6370

The slurmctld would segfault if we didn't check this.

Bug 6449.

What this code was used for was to try at the end of a job to see if
an association was there if there wasn't one there at the beginning of
the job.

From what we can tell the largest fallout here is a site wasn't
enforcing associations, then they do by scontrol reconfig and jobs
already running don't get an association id.  Since this was already
the case for any other job ran before hand this didn't seem like that
large of an issue.

What this does solve though is it allows you to release a job that was
held from a failed node since what was happening here was we got into
a state where if you ran

scontrol release 16862350_300
Job update not available right now, the DB index is being set, try again in a bit for job 16862350_300
slurm_suspend error: Job update not available right now, the DB index is being set, try again in a bit

This makes it so this state doesn't happen.

Bug 6340

already booted when slurmctld daemon is reconfigured.

Bug 6457

Bug 6468.

Bug 5513.

was recalculating if a new higher was larger than the previous but not
if existing highest decreased.

associations, partitions, qos, tres

Bug 6288

Bug 6286

gets out of sync easily when adding new sections.

Can be set by job_submit plugin to adjust a job's priority based on size
of job. It can also be updated with:
scontrol update jobid=<> adminpriofactor=#

Bug 6287

Modify GrpNodes limit to apply to unique nodes allocated (avoid double
counting nodes allocated to multiple jobs in the same QOS or association).

Bug 6458

Bug 6458

Lustre client stats moved to /sys/kernel/debug/lustre/llite with 2.12.

Bug 6385.

Co-authored-by: Tim Wickberg <tim@schedmd.com>

Add scontrol requeue option of "Incomplete" which will requeue jobs only if
    they failed to complete with an exit code of zero.
bug 6397

This is useful for job requeue.
bug 5693

Estimate the total count of GRES of a job at submit time.
This is used to reject jobs that will definitely exceed
a GRES limit rather than queuing them and rejecting them
when we finally determine their GRES count.
bug 6410