[mpich-discuss] multiple mpich2 jobs on a single node

[Anthony Chan]

Hi Bryan,

PBS variables provide info needed to mpdboot.
You should be able to use ${PBS_NODEFILE} to extact a count
for mpdboot.  e.g. assuming only 1 process per node.

NUM_NODES=`wc -l ${PBS_NODEFILE} | awk '{print $1}'`
mpdboot -n ${NUM_NODES} -f ${PBS_NODEFILE}

PS.  the awk is probably not necessary on newer version of wc.

A.Chan

----- "Bryan Putnam" <bfp at purdue.edu> wrote:

> On Thu, 1 May 2008, Rajeev Thakur wrote:
> 
> > In a PBS environment, you need not use MPD at all. You can simply
> use the
> > mpiexec written by Pete Wyckoff:
> http://www.osc.edu/~pw/mpiexec/index.php 
> 
> Hi Rajeev,
> 
> Yes we used to use that program when we used OpenPBS, but it doesn't
> work 
> with the latest PBSPro.
> 
> Bryan
> > 
> > Rajeev
> > 
> > > -----Original Message-----
> > > From: owner-mpich-discuss at mcs.anl.gov 
> > > [mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of Bryan
> Putnam
> > > Sent: Thursday, May 01, 2008 12:40 PM
> > > To: mpich-discuss at mcs.anl.gov
> > > Subject: RE: [mpich-discuss] multiple mpich2 jobs on a single
> node
> > > 
> > > On Thu, 1 May 2008, Rajeev Thakur wrote:
> > > 
> > > > Can you try running them with a single mpdboot?
> > > > 
> > > > Rajeev 
> > > 
> > > Rajeev,
> > > 
> > > Yes, it does work correctly when I use only one mpdboot as 
> > > you describe. 
> > > However, this is within a PBS environment, and the user doesn't
> know 
> > > whether or not the daemon is already running. 
> > > 
> > > Bryan 
> > > 
> > > 
> > >  > 
> > > > > -----Original Message-----
> > > > > From: owner-mpich-discuss at mcs.anl.gov 
> > > > > [mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of Bryan
> Putnam
> > > > > Sent: Thursday, May 01, 2008 7:52 AM
> > > > > To: mpich-discuss at mcs.anl.gov
> > > > > Subject: [mpich-discuss] multiple mpich2 jobs on a single
> node
> > > > > 
> > > > > Hi, perhaps this is an FAQ, but I can't seem to find it.
> > > > > 
> > > > > I've noticed that if I have two mpich2 jobs running on for 
> > > > > example, the same 8-processor node, submitted as
> > > > > 
> > > > > mpdboot
> > > > > mpirun -np 4 ./a.out
> > > > > 
> > > > > 
> > > > > mpdboot
> > > > > mpirun -np 4 ./a.out2
> > > > > 
> > > > > 
> > > > > then when one of the jobs completes, it kills all the mpd 
> > > > > processes along 
> > > > > with it, and the remaining job dies with message
> > > > > 
> > > > > job aborted; reason = mpd disappeared
> > > > > 
> > > > > 
> > > > > Is there an easy fix for this?
> > > > > 
> > > > > Thanks,
> > > > > Bryan
> > > > > 
> > > > > 
> > > > > --
> > > > > Bryan Putnam
> > > > > Rosen Center for Advanced Computing, Purdue University
> > > > > Young Hall (Rm. 519)
> > > > > 302 Wood Street
> > > > > West Lafayette, IN 47907-2108
> > > > > Ph 765-496-8225 Fax 765-494-0566
> > > > > bfp at purdue.edu
> > > > > http://www.purdue.edu/itap
> > > > > 
> > > > > 
> > > > > 
> > > > > 
> > > > 
> > > > 
> > > 
> > > 
> > > 
> > > 
> > 
> >