Hi One of our users found that array job steps are killed "randomly" and culprit seems to be epilog script.
https://github.com/SchedMD/slurm/blob/master/etc/slurm.epilog.clean problematic part is this: if [ $job_id -ne $SLURM_JOB_ID ] ; then exit 0 $job_id for array jobs have index counter and that comparison is invalid, e.g. 58183_4: integer expression expected epilog continues further and kills array job steps.. A bit hackish solution is to strip index counter, e.g. job_id=`echo $job_id | cut -f 1 -d_` Does audience have better solution? Best Regards, Tommi
