Fwd: Re: [Myricom help #17157] Myrinet errors
Susan Blackford <[email protected]> 05 Feb 2003 08:27:38 -0500
| Newsgroups | gmane.network.myrinet.general |
|---|---|
| Message-ID | <1044451660.412.355.camel@idefix> |
Oops, forgot to cc the Myrinet Mailing List on the reply. -----Forwarded Message----- > From: Myricom Technical Support <[email protected]> > To: Jeff Layton <[email protected]> > Cc: [email protected] > Subject: Re: [Myricom help #17157] [Myrinet] Myrinet errors > Date: 05 Feb 2003 08:25:21 -0500 > > Jeff, > > > > We're running our CFD code on a PIII/1.26 cluster with > > > Linux 7.3, gm-1.5.2, and MPICH-1.2.1..7b-5, the PGI > > > compilers (3.6 I think). Anyway, our codes have been working > > > in the past 12 months or so, but starting lately, we get error > > > messages that look like: > > > > > > > > > rcmd: socket: All ports in use > > > rcmd: socket: All ports in use > > > FATAL ERROR: timeout in initial synchronization sequence > > > Process aborting... > > > > > > Is this a Myrinet error or a rsh error (out of sockets)? > > It's a Linux configuration issue. > > Did you upgrade your Linux kernel recently? Maybe the number > of ports available was decreased (circumvent DoS attacks)? > > A search on www.google.com for "rcmd: socket: All ports in use" > found many hits for the "rcmd" manpage. > > socket: All ports in use > > All reserved ports in use. If a timeout occurs, check > whether the Internet Services are installed and inetd is > running. > > If you use MPICH-GM 1.2.4..x or later, you could spawn your > processes with MPD instead of rsh/ssh. This would avoid the > problem that you're having. > > Instructions for setting up and using MPD can be found on the Argonne > MPICH docs: > > http://www-unix.mcs.anl.gov/mpi/mpich/docs/mpichman-chp4mpd/node53.htm#Node55 > > Unlike with P4 where you need to specify ch_p4mpd as the device, > for MPICH-GM the ch_gm device will handle spawning with either > rsh/ssh or MPD. > > Details can be found in the README-mpich-gm, or the FAQ: > http://www.myri.com/scs/GM_FAQ.html#mpich-gm > > Hope this helps. > Susan > > -- > > ------------------------------------------ > Susan Blackford > Member of Technical Staff > Myricom Inc. > ------------------------------------------