order to interupt accept(), because this can fail if the authentication
plugin is bad or there are other communications problems. Use interrupt
instead.

  processed when job fails (e.g. due to SIGSEGV)

an assortment of other files in the installation lib/slurm directory
that we don't want to bother trying to open and report errors for.

nodes of the cluster.

and slurmctld (specify different configuration file).

SLURM credentials.

be expected.

than errno (which is not reset by the function).

MAX_SERVER_THREADS and increase that value from 50 to 60

results from apparent NFS synchronization delays).

name a prefix of "slurm_".

flag the required state saves. Perform the saves using synchronous I/O
from just one thread. Under heavy loads this results in much faster
responsiveness and lowers slurmctld's memory and CPU overhead considerably.

 o also show node state when -l is used with -R
 o Don't let environment var override command line (SINFO_FORMAT)

  of reasons and nodes with those reasons.

right away and treat failure as a fatal error.

write to.

    slurmctld startup
Create StateSaveLocation directory if changes via slurmctld reconfig

Properly support sinfo --state option value including "*" suffix.

  (which exhaust available memory).
Restructure slurmctld agent logic for higher throughput.