[MPICH] MPICH2 hangs on diskless SuSE 10.2 based cluster

[Shaun Q]

Rusty --

This I did. I can start mpd on one machine, manually connect it to another 
machine, do a ringtest, do a trace....  Still the same problem.

I am running the 1.0.5 release also, FYI.

Any ideas?

Thanks!
Shaun

>I recommend that you follow the steps in the User's Guide starting with 
>the use of mpdcheck to make sure both machines are properly configured 
and 
>can communicate with one another, and start the mpd's on these two 
>machines "by hand" instead of going immediately to the more scalable but 
>less informative mpdboot.
>
>Regards,
>Rusty
>

On Dec 20, 2006, at 6:33 AMDecember, Shaun Qualheim wrote:


     Hey everyone...


     I'm trying to figure out why I'm having an issue with running a job on 
multiple machines here.

     It worked fine with a 32-bit SuSE 9.3 based setup.


     I started it up with a 64-bit base distro and kernel...


     I can start up the process with:
     mpd &; mpiexec -n 2 process &
     and it starts with no issues...


     When I try doing that with 2 machines though...
     mpdboot -n 2 &; mpiexec -n 2 process &
     It hangs for about 5 minutes and then starts up.


     After it starts up and runs to completion, I can issue the same 
command again and it starts up right away.

     Any ideas of where to start here or what might be causing this issue?


     Thanks!
     Shaun