Closed andreas-el closed 2 months ago
The jobqueue does not terminate the jobs (seemingly) in earlier versions of komodo either.
2024.05.04
: UI blocked, jobs continue to run, stdout contains 'Job has already finished'
2024.05.00
: UI blocked, jobs continue to run, stdout contains 'Job has already finished'
2024.04.10
: UI blocked, jobs continue to run, stdout contains 'Job has already finished'
⚠️ 2024.03
: not released
2024.02.12
: UI blocked, jobs continue to run, stdout contains 'Job has already finished'
But what differs is that the "Experiment failed" dialog does not appear in these versions.
Terminated jobs around job ~10 done, still it continues.
[andrli@st-linrgs176 ~]$ bjobs
JOBID USER STAT QUEUE FROM_HOST EXEC_HOST JOB_NAME SUBMIT_TIME
656396 andrli RUN mr st-vgrid02 st-rsv17-17 *ly.ert-83 Jun 7 12:34
656401 andrli RUN mr st-vgrid02 st-rsv17-17 *ly.ert-88 Jun 7 12:34
656402 andrli RUN mr st-vgrid02 st-rsv17-17 *ly.ert-89 Jun 7 12:34
656403 andrli RUN mr st-vgrid02 st-rsv17-17 *ly.ert-90 Jun 7 12:34
656404 andrli PEND mr st-vgrid02 *ly.ert-91 Jun 7 12:34
656405 andrli PEND mr st-vgrid02 *ly.ert-92 Jun 7 12:34
This might be tied into the MAX_RUNNING
keyword. Testing was performed using the default value 2 (?), where we expect most users to set a rather high value for this, resulting in all jobs being submitted immediately.
We should also keep in mind that LSF10
was introduced not a long time ago.
Let's see if the same behaviour happens in scheduler. Update: it does not; ie. the jobs are terminated correctly.
We can keep this open until scheduler is enabled by default.
This has been a problem in the jobqueue for a long time, and given that it is fixed in the scheduler, we will not prioritize fixing this.
Using
2024.06.02
with scheduler set toFalse
on RGS.Also reproducible on
2024.02
,2024.04
&2024.05
releases, although these do not yield the error-dialog.Running poly-example Ensemble Experiment, and terminating jobs half-way through.
Job has already finished