In addition creating the individual log file for each array-job (ie, sbatch --output=job_%A_%a.out ...
), is there any way of getting a final report with all run jobs and see what ones have crashed? Something like that:
839594_1 COMPLETE
839594_2 FAIL
839594_3 COMPLETE
839594_4 COMPLETE
839594_5 COMPLETE
839594_6 FAIL
You should obtain the output you want with the sacct
command:
sacct -X -j 839594 -o jobid%-30,state
You can also add -n
to suppress the display of the column headers.
Note that all PENDING
jobs will be summarised into a single line like so:
839594_[10-20] PENDING
if tasks 10 to 20 are still pending.