[torqueusers] Warning against using Torque 2.1.5 ?

Garrick Staples garrick at clusterresources.com
Tue Oct 24 02:47:45 MDT 2006


On Tue, Oct 24, 2006 at 08:56:16AM +0200, Ole Holm Nielsen alleged:
> Since upgrading to Torque 2.1.5 we've seen a number of jobs
> mysteriously hang doing no work.  These jobs are all single-node
> Open-MPI jobs using 4 CPUs.  In /var/log/messages I see:
> 
> Oct 24 08:46:08 n057 pbs_mom: File exists (17) in open_std_out_err, Unable 
> to open standard output/error
> Oct 24 08:46:08 n057 pbs_mom: Inappropriate ioctl for device (25) in 
> start_process, cannot open job stderr/stdout files
> 
> and the MOM log says:
> 
> 10/24/2006 08:46:08;0001;   pbs_mom;Job;6614.audhumbla.fysik.dtu.dk;task 
> not started, 'orted', stdio setup failed (see syslog)
> 
> I guess that these problems may be related to Garrick's note in
> http://www.clusterresources.com/pipermail/torquedev/2006-October/000356.html
> 
> Perhaps it is necessary to issue a general warning against using
> Torque 2.1.5, and postpone upgrading until 2.1.6 is available ??

Current SVN trunk, 2.1-fixes, and 2.0-fixes should have everything
working correctly.  torque-2.1.6-snap.200610240247.tar.gz is ready to
download if anyone wants to test it.

It is well past my bedtime now.  If noone finds anything wrong, I'll do
new releases in about 12 hours (after lunch PST).



More information about the torqueusers mailing list