[gridengine users] funny output order

Reuti reuti at staff.uni-marburg.de
Fri Jul 19 10:18:07 UTC 2013


Am 19.07.2013 um 12:02 schrieb Ulrich Hiller:

> On 07/18/2013 09:13 PM, Reuti wrote:
> 
>> Is it possible, that several instances of this array job are running at the same time no one and the same host?
> 
> Yes it seems parts the job are randomly started several times.
> I found out that
> #$ -t 1-158
> could be the problem. When i set
> #$ -t 1-8
> (8 is the number of nodes) it runs only once as it should. A number <8
> means it does not run on all nodes, >8 results what i described.

Do you want to run an array job with several instances and make use of $SGE_TASK_ID?

Nevertheless you can restrict the number of array tasks running at a time with the `qsub` option "-tc".


> Yes, i have read the description of the -t option in the qsub manpage.
> But the question is why do i get such a result on such a simple job.
> I suppose i will get problems when i want to run a job on serveral CPUs
> on the same node.

Each job will need its own scratch directory which can be accessed by $TMPDIR. This is unique for each job (and task) and won't interfere with other jobs (or tasks) running on the same node. It would be best if you can adjust your jobscript to put this information in the designated location.

-- Reuti


> Did i misconfigure the gridengine?
> With kind regards, ulrich
> _______________________________________________
> users mailing list
> users at gridengine.org
> https://gridengine.org/mailman/listinfo/users




More information about the users mailing list