[gmx-users] Re: parallel run hangs (not crashed)

[chris.neale at utoronto.ca]

Can somebody post the particulars for how they submit a parallel run on their
system along with the brand/version of MPI they use?

Thanks everyone for recent help with parallel run hanging/crashing. I have
determined that it must be a problem with the way that I am submitting the
parallel job (i.e. not in my gromacs commands).

On David van der Spoel's suggestion I changes comm_mode to linear and
constraints to all-bonds. I am still running into the same problem.

On Mark Abraham's suggestion I verified that mpirun isn't a local wrapper.
However, he also suggested "mpirun may not play nicely with a single-processor
script that subsequently runs an MPI child process". I forwarded this to our lab
tech who was unsure what this meant or what the alternative was.

My original message was:
> I am running a system of 185K atoms. The structure is energy minimized and the
> dynamics run appears to be going smoothly until it just hangs. The job still
> exists on the first node, but none of the 4 nodes are doing any work and I don't
> get any error messages.