[torqueusers] always getting one node allocated independent of requested nodes

Walid walid.shaari at gmail.com
Fri Jun 13 11:42:57 MDT 2008


2008/3/18 <csamuel at vpac.org>:

>
> ----- "Walid" <walid.shaari at gmail.com> wrote:
>
> > We do see this with 2.1.6 from time to time (every 1, 3 months)
>
> Very bizzare, do people still see this with newer versions
> like 2.1.10 or 2.3 ?
>
> It sounds very much like a bug in pbs_sched and I suspect
> because it is considered an example implementation of a
> scheduler in C it doesn't tend to get a lot of love from
> the developers.
>
> Most people tend to need the extra functionality that
> Maui or Moab gives you.
>
> The fact that it went away when you switched to Maui
> seems to support that.
>
> cheers,
> Chris
>
I am using maui and torque 2.3.0 , and the event was triggered again, I
believe its the pbs_server that is responsible for this behaviour not
pbs_sched, as this time it was triggered when jobs were queued and did not
run, so a power user decided to use qrun to force them to run, and volia we
hit the bug again :(, for this and for the qhold, I am going to try 2.1.10
and see if we can stay long enough

regards

Walid
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20080613/09c6dad3/attachment.html


More information about the torqueusers mailing list