re-issuing it (reduces communications overhead).

KILL_JOB RPC to the slurmd, repeatedly if needed, but don't confirm it).
(gnats:326)

sent in a more timely fashion.

using node hostlist format.

for node registration, which is relatively rare).

very quickly due to a race condition (purge shared memory before completely
set up).

getting the configuration form slurmctld (which could be down).

primary and secondary controllers.

between primary and backup. The request has a brief window in which it can
abort and we want to decrease the likelyhood of that happening by retrying
less frequently when we know control is transistioning.

is still waiting to take control (rather than one long wait).

Slurmctld must be restarted for changes in these configuration parameters
to take effect. Previously these values would change, but there would be
no change in the already loaded plugin.

(jobcomp/none) in the configuration data structure.

data structure.

to take effect.

   force -gstabs if TotalView support is needed

 o Allow processing of UNSTABLE release for SLURM versioning

the natural consequence of a job ending immediately and before the shared
memory state information is completely established.
(gnats:321)

scontrol command to use it.

and/or job step(s) will have their resources de-allocated and be killed.
A resource allocation will not be release unless no job steps are active
for at least InactiveLimit seconds. DPCS jobs will be subject to this
forced de-allocation if they remain inactive for an extended period of
time, which can get SLURM and DPCS back in sync if DPCS does a cold-start.

were starting multiple jobs simultaneously and slurmd was not able
to respond to all of the requests without generating a message timeout.
(gnats:319)