Try running "lamclean" -- if your MPI application seg faults, shared
memory allocations may still be left around. lamclean (or lamhalt /
lamboot) should clear them out.
On Jul 20, 2007, at 8:12 AM, Miguel Ángel González Gisbert wrote:
> Hi all,
>
> I am currently developping a paralell application with LAM/MPI and
> everything used to be fine. However, since yesterday I cannot work any
> more with the system.
>
> I tried to isolate the problem so I tried to check if MPI works with a
> simple paralell "Hello world!". I'm getting the following error:
>
> magonzalez_at_baobab:~/StageVerano/MultMatrices/MPI/Ver1$ mpirun -np
> 3 ./hello
> ----------------------------------------------------------------------
> -------
> The selected RPI failed to initialize during MPI_INIT. This is a
> fatal error; I must abort.
>
> This occurred on host n2 (n1).
> The PID of failed process was 12909 (MPI_COMM_WORLD rank: 2)
>
> I also tried with only 2 cluster nodes and there's is no problem
> but... there isn't any output! (before THERE WAS an output):
>
> magonzalez_at_baobab:~/StageVerano/MultMatrices/MPI/Ver1$ mpirun -np
> 2 ./hello
> magonzalez_at_baobab:~/StageVerano/MultMatrices/MPI/Ver1$
>
> I have searched the answer for my problem into the list and the
> Internet but I found no satisfactory solutions.
>
> I will thank any help :),
>
> Miguel Ángel González Gisbert
> FACULTAD DE INFORMÁTICA
> UNIVERSIDAD POLITÉCNICA DE MADRID
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
--
Jeff Squyres
Cisco Systems
|
