i found a solution in
http://www.gridengine.info/2006/02/14/grouping-jobs-to-nodes-via-wildcard-pes/
the following is more suitable for my needs :
create a host group for each subcluster – group1 group2
create pe for each subcluster fillup_group1 fillup_group2
associate the pes to queue
create resource quota rule
limit pes {fillup_group1} hosts {!group1} to slots = 0
limit pes {fillup_group2} hosts {!group2} to slots = 0
On Mon, Oct 29, 2012 at 9:36 PM, <[email protected]> wrote:
>
>
> In the message dated: Mon, 29 Oct 2012 14:24:14 +0200,
> The pithy ruminations from Oren Mustaki on
> <[gridengine users] preventing mix between 2 hosts group> were:
> =>
> => Hi everybody
> =>
> => i have 2 groups of 36 nodes each. each group is connected via Infiniband
> => switch.
> => the 2 Infiniband switches are *not* connected.
> =>
> => the groups are:
> => par.q@@group1 , par.q@@group2
> =>
> => the -q line:
> => #$ -q par.q@@group1,par.q@@group2
> =>
> => once in a while a job is trying to run on nodes from both groups which
> => result in failure.
> =>
> => is there a way to allow running on par.q@@group1 or par.q@@group2 but
> never
> => on both ?
>
> You could:
>
> put both groups into the same queue
>
> assign a boolean attribute to each group (for example,
> each server in "group1" has the attribute "group1")
>
> The submission would look like:
>
> qsub -q combined_groups -l group1 myjob
>
> or
>
> qsub -q combined_groups -l group2 myjob
>
>
> Without specifying the resource requirement, the job would be free to run
> in
> either group, a benefit for things like single-slot jobs, as in:
>
>
> Mark
>
>
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
