On Wed, Sep 28, 2005 at 12:47:11PM +0200 or thereabouts, Maarten Litmaath, CERN wrote:
> On Wed, 28 Sep 2005, Antun Balaz wrote:
>
> > Hi,
> > I am using torque/maui combination. So if poweroff is imminent, can I apply
> > the following procedure:
> >
> > [say we have only one queue (thequeue), and one running job in it (with job
> > ID equals JOBID)]
> >
> > CE:
> > qdisable thequeue
> > mjobctl -R JOBID
> >
> > All nodes:
> > restart (including poweroff)
> >
> > CE:
> > qenable thequeue
> >
> > Will the job JOBID survive and start execution again from the beginning, and
> > will the user be able to fetch the output using edg-job-get-output once tha
> > job is completed?
>
> Do _not_ enable such functionality: many VOs explicitly set the RB resubmission
> count to zero because they _cannot_ handle a job possibly getting run twice.
> For them it is better to have the jobs fail.
> Power cuts are a fact of life, do not worry too much about them.
All jobs that I see submitted from the RB are marked as not rerunable
with in torque. i.e. the above won't work.
Steve
--
Steve Traylen
[log in to unmask]
http://www.gridpp.ac.uk/
|