On 04/03/2013 04:49 AM, Reuti wrote:
Am 03.04.2013 um 08:23 schrieb Joseph Farran:
Howdy.
Using GE 8.1.2. I have two jobs which suspended correctly via Grid Engine
subordinate queue.
I am however trying to force the scheduler to resume ( un-suspend ) the
suspended jobs with no success:
$ qstat | grep compute-14-18
288279 0.50000 MakeSummar juser S 04/02/2013 16:00:43
[email protected] 1 69788
288279 0.50000 MakeSummar juser S 04/02/2013 16:00:43
[email protected] 1 69827
289206 0.33333 augustus_s muser r 04/02/2013 18:24:16
[email protected] 32
289278 0.33333 monti_augu muser r 04/02/2013 21:08:48
[email protected] 32
Is this defined as slotwise subordination or in the traditional way (there will
be an additional S at the end of each cluster queue in `qstat -f` in the latter
case)?
Hi Reuti.
I am using traditional queue sub suspension:
$ qconf -sq sam | fgrep sub
subordinate_list free64=1, abio=1
The jobs were killed so I cannot do the qstat -f on it now.
Suspend by subordination and by `qmod` are different things.
$ qmod -usj 288279.69788 -f
root - forced enabling of job-array task 288279.69788
$ qstat | grep compute-14-18
288279 0.50000 MakeSummar juser S 04/02/2013 16:00:43
[email protected] 1 69788
288279 0.50000 MakeSummar juser S 04/02/2013 16:00:43
[email protected] 1 69827
289206 0.33333 augustus_s muser r 04/02/2013 18:24:16
[email protected] 32
289278 0.33333 monti_augu muser r 04/02/2013 21:08:48
[email protected] 32
Is there a way to force these the two jobs ( 288279.69788 and 288279.69827 ) to
resume?
You could ignore the output and send a `kill -cont` on the node to the complete
process group of the jobs. The output maybe still wrong then, but at least the
jobs may continue.
I think I tried that with no success, so I am wondering now if the jobs were
just stuck.
IIRC you had the opposite problem in the past: jobs were suspended but
continued anyway in the process listing.
Yes. That other issue seems to have subsided for now, but yes I was having
jobs continue to run while Grid Engine had them as suspended.
Best,
Joseph
-- Reuti
Thanks,
Joseph
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
