Hi,

(this is reviving a year old thread, I came across it by chance)

To solve issues like what you described, as well as some related parallel
allocation jobs, this qtop tool was written: https://github.com/qtop/qtop
I'm not sure how compatible it is with a given version of SGE, but it has
been used daily for years on some dozen SGE clusters, without trouble.

Hope it helps someone out there.

cheers,
F.

p.s. happy to receive feedback also via private email or, on github.

On Wed, 12 Dec 2018 at 21:54, Dj Merrill <s...@deej.net> wrote:

> Thank you for the quick response.
>
> I've setup three groups, @cpu1hosts, @cpu2hosts, and @cpu5hosts.
>
> The @cpu1hosts consists of the node1* nodes, @cpu2hosts the node2*
> nodes, and @cpu5hosts the node5* nodes.
>
> In the default queue I have:
>
> hostlist              @cpu1hosts @cpu2hosts @cpu5hosts
>
> seq_no                1,[@cpu1hosts=10],[@cpu2hosts=20],[@cpu5hosts=50]
>
>
> In addition to these changes, I had to do "qconf -msconf" and change
> "queue_sort_method" from "load" to "seqno".
>
> I believe this is now doing what we want.
>
> Thank you!  :-)
>
> -Dj
>
>
>
> On 12/12/18 2:18 PM, Kamel Mazouzi wrote:
> > Hi,
> >
> > It's about seq_no parameter in queue definition.
> >
> > Assume that you have 2 groups of nodes: node10all and node50all
> >
> > Add this to your queue definition:
> > seq_no                1,[@node10all=50],[@node50all=100]
> >
> > New jobs will start running on nodes node50all
> >
> > Best regards
> >
> >
> >
> >
> >
> >
> > On Wed, Dec 12, 2018 at 8:05 PM Dj Merrill <s...@deej.net
> > <mailto:s...@deej.net>> wrote:
> >
> >             I seem to recall reading about this in the past, but I can't
> >     locate the
> >     reference so I'm hoping someone might be able to help.
> >
> >             We have some older nodes on our SGE system, starting at
> >     node50.  We
> >     added some new nodes to the system, starting at node10.  With a
> >     completely empty Grid (no jobs), we would like newly submitted jobs
> to
> >     start running on node10 (ignoring for the moment any special resource
> >     requests).  Instead, the jobs start on node50, which is the "oldest"
> >     node on the Grid in terms of how long it has been a part of the Grid.
> >
> >             Is there an easy way to specify a preferred usage order, or
> >     will we
> >     have to remove all of the old nodes from the Grid, then re-add them
> as
> >     new nodes?
> >
> >     Thanks,
> >
> >     -Dj
> >     _______________________________________________
> >     users mailing list
> >     users@gridengine.org <mailto:users@gridengine.org>
> >     https://gridengine.org/mailman/listinfo/users
> >
> _______________________________________________
> users mailing list
> users@gridengine.org
> https://gridengine.org/mailman/listinfo/users
>


-- 
echo "sysadmin know better bash than english"|sed s/min/mins/ \
  | sed 's/better bash/bash better/' # signal detected in a CERN forum
_______________________________________________
users mailing list
users@gridengine.org
https://gridengine.org/mailman/listinfo/users

Reply via email to