[gmx-users] mpi_run issue

[Mark Abraham]

----- Original Message -----
From: #ZHAO LINA# <ZHAO0139 at ntu.edu.sg>
Date: Friday, January 29, 2010 1:11
Subject: [gmx-users] mpi_run issue
To: gmx-users at gromacs.org

> Hi,
> 
> I tried to run the command
> mdrun -deffnm em
> on cluster,
> blow is the script, but I worried that it might be some place 
> wrong, seems no files came out after several hours running
> Is anyone please help me to modify it, and how can know the 
> process of its running in the cluster.

We can't know the best procedure for your cluster because we haven't read its documentation or talked to its admins... (hint, hint)

> Thanks and best regards,
> 
> Elsa
> 
> #!/bin/bash
> #
> ## Specify the job name
> #PBS -N emMDrunSecondOne
> 
> ## Join the standard error and the standard output into 1 file output
> #PBS -j oe
> #PBS -V
> 
> ## To run on 16 cpus
> #PBS -l nodes=2:ppn=8
> 
> PBS_O_WORKDIR=${HOME}/Jobs/1A7W/tryScript
> [ -d ${PBS_O_WORKDIR}] || exit 1

This looks wrong. PBS usually has flags that let you specify running the job in directory from which submission occured - use them.
 
> cd $PBS_O_WORKDIR
> NCPUS=`cat $PBS_NODEFILE | wc -l`
> echo $NCPUS

If PBS is defining $PBS_NODEFILE to point to some file it creates, then that file won't be local to the directory you've arbitrarily changed to, so NCPUS won't get defined suitably, so mpirun won't work.

I'd suggest you start again from some example script that is known to work on your cluster and modify only the mpirun line.

Mark

> ## program to run
> /usr/local/openmpi/bin/mpirun -np $NCPUS mdrun_mpi -deffnm em 
> exit 0
> > -- 
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search 
> before posting!
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php