[slurm-users] Re: orted errors

What's the best way to troubleshoot this when orted fails but doesn't give any sort of error to indicate what the root cause of the failure might be? And I also can't predictably induce the failure, just have to wait until it randomly chokes.

You can try increasing the Open MPI verbosity — generally and module-specific. That's often how I am able to notice what's wrong under the hood with Open MPI. Use the `ompi_info` command to check for all "verbose" parameters:

$ ompi_info --level 9 --all | grep _verbose