Max_run_subjobs vs max_run nuances

Curious what the exact nuances is between max_run_subjobs vs max_run ? I think they can both be specified as a queue attribute (?) so is there a nuance here, cause they both server to limit a number of subjobs running. Is there a queuing difference or priority? Does max_run_subjobs on the cli override the queue attribute? Actually, and is there a way to specify a sep. priority for subjobs? (they are regular jobs, right? So thinking not.)

Thanks

max_run_subjobs : to limit number of subjobs running at the same time
qsub -J 1-100000%1000 pbsscript.sh
qsub -J 1-10000%<max_run_subjobs> pbsscript.sh
Refer: elim_on_subjobs

max_run : maximum number of jobs allowed to be running
With respect to the job array: Refer: elim_on_subjobs , this has some implications on understanding on count subjob as single jobs or count array job as single job which is default behaviour.

All the subjobs of a job array has the same scheduling priority.
Individual subjobs priority cannot be set.

Not regular jobs in terms of assigning priority.
Refer: elim_on_subjobs affects max_queued limits

Ty!

For reference, the bit I found pertinent to me from the elm_on_subjobs

“specify whether the max_queued limit attribute counts each array job as a single job, or counts each subjob as
a single job via the elim_on_subjobs server attribute. Set this to True to have each subjob count as a single job.”

For others.

Def do NOT want it counting against max_queued.

1 Like