[torqueusers] What is cartwire.ncsa.uiuc.edu?
garrick at usc.edu
Tue Oct 18 12:59:22 MDT 2005
On Tue, Oct 18, 2005 at 10:48:01AM +0200, Lorenzo Campo alleged:
> no, I'm using the source of torque 1.2.0p6, downloaded and compiled on my
> machines following the quickstart guide (and compiling with --with-scp
> option). I'm using the pbs_sched scheduler, no RPM or other binary at all.
> There is some issue in running jobs with more than one compute node?
TORQUE routinely runs jobs with hundreds of nodes probably every second
of every day.
Looking at those error messages again, I think this is in your MPI
stuff, not in TORQUE.
Which MPI implementation are you using? How are you running helloMPI?
I'm guessing you might be using MPICH with the p4 device, in which case,
make sure you run 'mpirun' with some variation of:
mpirun -machinefile $PBS_NODEFILE -np `wc -l < $PBS_NODEFILE`
Alternatively, mpiexec from http://www.osc.edu/~pw/mpiexec/ integrates
very nicely with TORQUE and is widely used and recommended.
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/20051018/4be31e7d/attachment.bin
More information about the torqueusers