[gridengine users] stopping all jobs on a cluster queue at 0100 everyday.

Oren Mustaki mustaki at phys.huji.ac.il
Sun Nov 4 13:49:15 UTC 2012


ill use a "qstat | awk | xargs qdel" crontab line to kill the jobs .
i tough tit would be easier from the queue itself

Thanks

0ren

On Sat, Nov 3, 2012 at 9:52 PM, Reuti <reuti at staff.uni-marburg.de> wrote:

> Hi,
>
> Am 03.11.2012 um 15:55 schrieb Oren Mustaki:
>
> > Is there a way to stop all jobs  ( usually qrsh ) on a cluster queue at
> a certain clock time rather then cpu time.
> > for example all jobs on shell.q will be deleted at 0100.
> > it can be done using bash or other scripting but i want to know if it
> can be done using SGE builtin features .
>
> in principle: yes. This would need the setup of a calendar in SGE to
> suspend the queue at a certain time, and a redefined suspend-method in the
> queue could then kill the jobs instead of suspending them. As long as
> "max_reservations" in SGE configuration is set to 0, this will work,
>
> But: jobs will either use the default run-time from the SGE configuration
> or the one requested at submission time. And as a result, the job won't
> start if SGE detects that they won't finish before the suspension in case
> you need max_reservation being set to some value different from 0 to take
> place.
>
> -- Reuti
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gridengine.org/pipermail/users/attachments/20121104/a2b42edd/attachment.html>


More information about the users mailing list