Moin Thomas,we are using the slurm epilog to append some informational data to STDOUT after the job finished.
in slurm.conf : Epilog=/etc/slurm/epilog.sh in epilog.sh (with root privileges) s.th. like this: JOBINFO=`scontrol show job $SLURM_JOB_ID`SLURM_STDOUT=$(echo $JOBINFO | grep -o "StdOut=[^ ]*" | awk -F= '{print $2}')
if [[ -f ${SLURM_STDOUT} ]]; then echo "REPORT" | tee -a ${SLURM_STDOUT}; fi
In such a way you can give any information from scontrol or sacct or
whatever metrics you calculated while the job was running.
Hope this helps, Hendryk On 10.08.2015 20:31, Thomas Orgis wrote:
Am Mon, 10 Aug 2015 10:51:58 -0700 schrieb "Zentz, Scott C." <[email protected]>:Hey Marcin, From what it looks like, the AccountingStorageLoc sounds like an option for storing job information for every job in a location specified by an admin. I am looking for a way for the end users who are not interested in receiving email and instead logging the time their job took to process. Will the AccountingStorageLoc accomplish that?You can lookup the data later with sacct. But I also would like an easy way to print out such info in a file related to the job at exit. We used to have such a report appended to the STDERR output file on another cluster: ======================== Job performance report ======================== hosts: 1 processes: 8 CPU time: 00:03:51 wallclock time: 00:00:52 total CPU usage: 56 % total physical memory: 0.346 GiB phys. mem. / host (avg.): 0.346 GiB total virtual memory: 5.164 GiB virt. mem. / host (avg.): 5.164 GiB ====================== End of performance report ======================= I used this to also diagnose very suboptimal CPU usage (lots of I/O to the wrong place) and add a suggestive message to the report. I'd like to do the same with Slurm, but so far I didn't see a way to write that kind of info to the job output file automatically. Alrighty then, Thomas
smime.p7s
Description: S/MIME Cryptographic Signature
