Am 06.02.2012 um 22:44 schrieb Reuti:
> <snip>
> As the parent jobs are not doing any work, a special parent.q would do which
> has to be requested by a forced boolean complex, so that only parent jobs can
> get in. You could even set a h_cpu limit on this queue to avoid abuse - jobs
> abusing this queue would get killed after 5 minutes or so. The overall slot
> count used in this cluster queue you can limit in an RQS.
I didn't mention it: you oversubscribe by intention. In case you have further
queues on each host, the RQS for limiting the overall slot count will need two
entries:
limit name parents queues parent.q hosts {*} to slots=1
limit name slots hosts {*} to slots=$num_proc
The first matching rule will end the RQS. The second RQS for the parent.q
overall use it like usual:
limit name parents queues parent.q to slots=17
Please adjust the numbers to your needs. You could even run all parents on one
machine, but I would spread it around.
-- Reuti
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
