Evening I have built a nice little cgi script which helps users monitor their jobs. We have noticed that when an array has a non-1 index, squeue returns a stupid number of array tasks... > tab=" " > squeue -j 21075999 PARTITION PRIORITY NAME USER ST TIME TIME_LEFT NODES NODELIST(REASON JOBID XXXXXXXXXXX 200 zzzzzzzzzzzzzzzzzzzz yyyyyyyy PD 0:00 21:00 1 (Resources) 21075999_[22655-22999:8] XXXXXXXXXXX 200 zzzzzzzzzzzzzzzzzzzz yyyyyyyy R 13:14 7:46 1 node1 21075999_22639 XXXXXXXXXXX 200 zzzzzzzzzzzzzzzzzzzz yyyyyyyy R 8:00 13:00 1 node2 21075999_22647 > squeue -h -r -a -j 21075999 -o "%t${tab}%F${tab}%K${tab}%P${tab}%Q${tab}%u${tab}%j${tab}%k${tab}%Z${tab}%M" | wc -l 207346 and if I look at the output of the command, we have array tasks larger than 22999! PD 21075999 22996 XXXXXXXXXXX 200 yyyyyyyy zzzzzzzzzzzzzzzzzzzz aaaa/bbbb/workflow.job aaaa/bbbb/ 0:00 PD 21075999 22997 XXXXXXXXXXX 200 yyyyyyyy zzzzzzzzzzzzzzzzzzzz aaaa/bbbb/workflow.job aaaa/bbbb/ 0:00 PD 21075999 22998 XXXXXXXXXXX 200 yyyyyyyy zzzzzzzzzzzzzzzzzzzz aaaa/bbbb/workflow.job aaaa/bbbb/ 0:00 PD 21075999 22999 XXXXXXXXXXX 200 yyyyyyyy zzzzzzzzzzzzzzzzzzzz aaaa/bbbb/workflow.job aaaa/bbbb/ 0:00 PD 21075999 23000 XXXXXXXXXXX 200 yyyyyyyy zzzzzzzzzzzzzzzzzzzz aaaa/bbbb/workflow.job aaaa/bbbb/ 0:00 PD 21075999 23001 XXXXXXXXXXX 200 yyyyyyyy zzzzzzzzzzzzzzzzzzzz aaaa/bbbb/workflow.job aaaa/bbbb/ 0:00 PD 21075999 23002 XXXXXXXXXXX 200 yyyyyyyy zzzzzzzzzzzzzzzzzzzz aaaa/bbbb/workflow.job aaaa/bbbb/ 0:00 (sorry for anonymising it)
Hi, This looks like old solved issues, commit cdf2587850 fixes this problem. The patch is available as https://github.com/SchedMD/slurm/commit/cdf258785034f825234b325.patch You can manually apply this patch. Dominik
Sigh, thanks. I guess we really should get upgraded to a more recent version :) (In reply to Dominik Bartkiewicz from comment #1) > Hi, > > This looks like old solved issues, > commit cdf2587850 fixes this problem. > The patch is available as > https://github.com/SchedMD/slurm/commit/cdf258785034f825234b325.patch > You can manually apply this patch. > > Dominik
Marking resolved, please reopen if there were any further questions. Dominik