Hi Bill, Thanks for reporting this. Sounds like something we need to handle in the Hydra name resolution code. I've created a ticket in Trac for it. http://trac.mpich.org/projects/mpich/ticket/2263 Ken On 05/28/2015 10:44 PM, Bill Broadley wrote:

I tried a quite a few variations on ubuntu's slurm, slurm-14.11.7, and numerous build options for mpich.

Turns out the ubuntu incompatibility with mpich is because in /etc/hosts the hostname typically resolves to 127.0.1.1. Thus when hydra grabs IPs for both hosts it ends up with the wrong IP.

If I remove the line in /etc/hosts with the hostname and 127.0.1.1 it just works. Might be worth having mpich warn about using a 127.0.1.1 IP, at least when using more than one node. I poured over the verbose logs and hydra debug logs without noticing the problem.

OpenMPI doesn't have this problem for whatever reason.

_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss