I have determined that it is the sm btl that is hanging, and that it has something to do with locality. I’ll have to dig deeper in the morning.
For now, I have reverted the commit that seems to be causing the problem. Sorry for the trouble. Ralph > On Sep 11, 2015, at 1:28 AM, Ralph Castain <r...@open-mpi.org> wrote: > > Yo folks > > I just discovered that something is borked in the master - we are hanging on > multi-node startup. I’m unsure where it crept into the system, but sometime > in the last 24 hours, so I’ll try to figure it out. Looks like it is > something in PMIx, but I haven’t confirmed it yet. > > Just a heads-up > Ralph >
