[torqueusers] Output capture problem
garrick at usc.edu
Thu Jun 30 21:02:44 MDT 2005
On Thu, Jun 30, 2005 at 07:32:28PM -0700, Garrick Staples alleged:
> On Tue, Jun 28, 2005 at 10:50:59AM +0200, Enrico Morelli alleged:
> > Dear all,
> > I have a cluster with torque torque-1.2.0p2 and maui 3.2.6p13 and I'm
> > trying to catch the calculations output.
> > If I use a job script with the directives like
> > #PBS -j oe
> > #PBS -o job.out
> > the file job.out isn't create.
> This is telling pbs to merge stdout into the stderr stream, but defines an
> output file for the non-existant stdout stream. You might actually want "-j
I just double checked this stuff and my understanding of "oe" was backwards.
Those options will cause a single stdout file to be created in $PBS_HOME/spool
on the execution which will be copied to job.out after the job completes.
If that isn't working, then rcp or ssh isn't setup correctly, or you need to
use the $usecp option in mom's config.
> > The only way to have some output is to use the directives:
> > #PBS -j oe
> > #PBS -k oe
> > that creates a file script.oNJOB in the home directory.
> -k tells pbs to create the files directly in your home directory. IIRC, it
> overrides -o and -e, but I'd have to check the code again to be sure.
> > My questions are:
> > 1) Why the -o doesn't works?
> Just for fun, try it again without -j or -k.
> > 2) Why the option -k write the output files in the home directory also
> > if the "qstat -f" command displays
> > Error_Path = athlon.cluster.net:/home_n10/zntA/DYANA/PBSTEST/run.e80
> > Output_Path = athlon.cluster.net:/home_n10/zntA/DYANA/PBSTEST/run.o80
> I'm pretty sure -k overrides -o and -e.
I confirmed that -k does override -o and -e. And I also found that this
plainly stated in the qsub manpage :)
> > 3) Why the writing of the output in the files require some times,
> > there is a flag to reduce the caching?
> Depending on these various options, pbs_mom will either write the output to
> files in $PBS_HOME/spool/ and then copy files to their final destination after
> the job ends, or write the output directly to the named output files. I'm not
> sure which scenerio is causing you to ask this question. If it's the later,
> then it's just NFS behaving normally.
> And quite frankly, I'm not sure all of these options are all working correctly.
> Garrick Staples, Linux/HPCC Administrator
> University of Southern California
> torqueusers mailing list
> torqueusers at supercluster.org
Garrick Staples, Linux/HPCC Administrator
University of Southern California
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torqueusers/attachments/20050630/d498d0bd/attachment.bin
More information about the torqueusers