I working on a problem related to enforcement of a MaxSubmit limit for a
QOS. While doing some investigation I saw the following statement in the
srun man documentation regarding the '-t' option:
-t, --time=<time>
Set a limit on the total run time of the job or job step.
If
the requested time limit for a job exceeds the partition's
time
limit, the job will be left in a PENDING state (possibly
indefi-
nitely).
I suspect there may be some correlation between what is being seen as the
MaxSubmit limit being exceeded and jobs that have been submitted with '-t'
values larger than the QOS limit. The user
checked the number of jobs submitted with an 'sacct' command and if there
aren't accounting record yet for these PENDING jobs they wouldn't be seen
using this command
I have tried unsuccessfully to reproduce this situation because I can't
get the jobs to be left in the PENDING state. Can anyone please tell me
under what conditions the job will be left in the PENDING
state rather than being cancelled?
Best Regards,
Bill
