Thanks, I looked at this version of fsl_sub. It is similar to ours but does use the afterokarray syntax that should theoretically solve the problem with dependence on an entire array. Unfortunately, this syntax does not behave as expected on my system.
I do not see anything in this code that handles the dependence on a job that has completed and left the grid. This is not a problem for many calls to feat but does occur for me sometimes in event models or Melodic when the list of dependencies is very long, taking several minutes to submit.
Kathy
--------------------------------------------------------
Hi,
I think a while back that Matt Glasser posted the 'fsl_sub' that we use on
our TORQUE system. For convenience, I'm again attaching the 'fsl_sub'
that we are using currently with FSL 5.0.6.
It is provided solely as is, in the hope that it may be beneficial to
others using TORQUE. I can not provide any support if it doesn't work on
your particular system.
Good luck,
-MH
--
Michael Harms, Ph.D.
-----------------------------------------------------------
Conte Center for the Neuroscience of Mental Disorders
Washington University School of Medicine
Department of Psychiatry, Box 8134
660 South Euclid Ave. Tel: 314-747-6173
St. Louis, MO 63110 Email: [log in to unmask]
On 2/19/14 12:50 PM, "Kathy Pearson" <[log in to unmask]> wrote:
>Michael,
>
>So far I have only a work-around, no real solution, for these problems
>noted with redhat torque and moab:
> (a) if a job completes, no more dependencies may be added to it with
>qsub -W
> (b) the array dependence jid[N] may need to be expanded to jid[1]
>jid[2] ... jid[N] --- or perhaps replace afterok with afterokarray
>
>For (a), I have simply used the qsub -h option to hold all jobs, then
>called qrls to release them after feat finishes queueing all of the
>tasks. I am unsure of how to incorporate this idea into fsl_sub. Using
>the qsub -h option is merely an expedient to make our grid behave more
>like SGE where I never saw this problem.
>
>The (b) question could likely be resolved within fsl_sub, but I cannot
>get the afterokarray syntax to work as documented on my system and
>stringing out jid[1]:jid[2] ... jid[91] as I have done certainly seems
>inelegant.
>
>Kathy
>
>--------------------------------------------------------
>Hi Kathy and Henk,
>
>could I maybe interest you in collaborating on github (or any other scm
>resource) for this particular matter?
>
> https://github.com/octomike/fsl_overlay
>
>I would really love to see your changes and add it to our torque/fsl_sub
>overlay !
>
>cheers,
>Michael
|