UNIX for Dummies Questions & Answers

Hi everyone -

I'm trying to start a run of LAMMPS (which takes an input file) with mpiexec on multiple processors. I want to combine this with nohup so that I can log off the node. I've used the syntax below successfully on one cluster, but it doesn't work on the new one I'm using. Any suggestions would be helpful.

First, I have a script:

Code:

#!/bin/bash
mpiexec -np 8 /home/foo/lmp_openmpi < /home/foo/inputfile

Now, I call that script with nohup via:

Code:

nohup sh script > nohup.out &

This starts the lammps job just fine, on 8 processors. However, when logging out of the node, the jobs die. Checking the nohup.out, I get the final line:
"mpiexec noticed that process rank 1 with PID 37372 on node whatever exited on signal 1 (Hangup)."

Isn't that what nohup is supposed to be preventing? I've seen the redirecting to /dev/null thing mentioned around the web, but I'm either using it wrong, or it's ineffective. Help!?!