dolfin not found under slurm sbatch

Question

Hello,

I've recently installed fenics on a cluster. I can now execute my code with "python myscript.py" and "mpirun python myscript.py" and a bash script containing any of them. However, when I try to run this bash script through "sbatch -A myaccount -p myqueue bashscript", python is not able to find dolfin at all. I always get:

Traceback (most recent call last):
File "kk.py", line 1, in
from dolfin import *
ImportError: No module named dolfin

I tried exporting all environment variables (especially those from fenics.conf) through sbatch and mpirun, but no luck. I know this may not be a directly related question to fenics issues, but I'm writing in case anyone had a simillar issue (it's my last step to have a fully functioning fenics on a cluster)

Thank you very much for your help.

Comments

Did you set any environment variables in your bash script?
Use env to list them all before calling python.

#!/bin/bash

export PYTHONPATH=/my/python/lib:$PYTHONPATH
env
mpirun -n 4 python example.py 

Answer 1

Hello,

Yes I tried exporting all env variables, but it was again a cluster configuratoin issue. My cluster admin fixed it. Just in case someone has a similar issue: the filesystem where I compiled and installed fenics is a local filesystem - I couldn't do it on my home dir which is shared among all nodes from the cluster because of the quota limit. The cluster admin just moved the fenics installation to a location accesible to all the nodes and it's now working.

Thank you for your help.