Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: Re: [OMPI users] Issue with : mca_oob_tcp_peer_recv_connect_ack on SGI Altix
From: Jeff Squyres (jsquyres_at_[hidden])
Date: 2010-12-16 07:04:36

On Dec 16, 2010, at 3:29 AM, Gilbert Grosdidier wrote:

>> Does this problem *always* happen, or does it only happen once in a great while?
> gg= No, this problem happens rather often, almost every other time.
> Seems to happen more often as the number of cores increases.

Well that's a bummer -- it seems to indicate that this may be a problem in OMPI.

Are you running multiple OMPI jobs concurrently? More specifically, are you starting multiple jobs on the same machine more-or-less at the same time? I'm wondering if our TCP startup mechanism is somehow accidentally getting the TCP ports from a different job. I can't imagine how that would be happening, but...

> gg= Is there a way with the current code, to direct OpenMPI to use a restricted range of TCP ports,
> that I can choose at launch time ?

Yes. In OMPI v1.4, there's 2 MCA params:

oob_tcp_port_min_v4 (default: 0)
oob_tcp_port_range_v4 (default: 65536)

Try setting these values to mutually exclusive ranges for each of your jobs and see if that fixes the problem. Keep in mind that user-level ports start at 1024, so your lowest range might as well start at 1024.

Jeff Squyres
For corporate legal information go to: