[torqueusers] lamboot, mpirun, pbs nodefile, and torque

Garrick Staples garrick at usc.edu
Tue Jul 1 12:27:17 MDT 2008


On Tue, Jul 01, 2008 at 10:51:04AM -0400, Glen Beane alleged:
> On Mon, Jun 30, 2008 at 3:53 PM, <bkovacs at fusiongeo.com> wrote:
> 
> > Hi,
> > I use lamboot with mpi to run my jobs. With torque and pbs I understand I
> > need PBS_NODEFILE to get a list of nodes for the mpirun to use. I am using
> > a bash script to keep all the variables and job info. The thing is that I
> > need the jobs to run from the head node not the compute node that the
> > PBS_NODEFILE is stored on. How do I get all the jobs to run from the head
> > node and have it able to have the PBS_NODEFILE on the head node?
> > Thank you,
> > BB
> 
> 
> the PBS_NODEFILE is created at job run time by the pbs_mom.  It will be on
> the compute nodes, not the head node, but if you are using LAM then
> _DO_NOT_USE_ PBS_NODEFILE!!  Compile LAM with TM support.  Then do this in
> your TORQUE job script

Glen, I think you missed the question there somewhere (but I don't really
understand the question either).

BB, It seems non-sensical to want torque to run jobs on a set of nodes, but
actually want the job to run somewhere else.  I think you need to rethink your
workflow.

-- 
Garrick Staples, GNU/Linux HPCC SysAdmin
University of Southern California

Please avoid sending me Word or PowerPoint attachments.
See http://www.gnu.org/philosophy/no-word-attachments.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torqueusers/attachments/20080701/7d21de8f/attachment.bin


More information about the torqueusers mailing list