Hello Simon,
On Thu, 15 Jun 2006, Simon Prunet wrote:
> Dear LAM/MPIers,
>
> We have here a cluster of opterons, running Suse
> with OpenPBS and Lam-mpi (7.1.1).
>
> It seems that the combination of openpbs queuing
> and lam mpirun result in a stack size limit of 8192 kb,
> which for some fortran codes can result in segfaults.
>
> I have checked that openpbs alone, or mpirun alone
> do not result in this limit, but only the combination of
> both...
>
> Has anyone encountered this problem ?
We encountered a similar problem, try setting the stacksize in the pbs mom startup script to the desired value. In our case we added `ulimit -s unlimited` to the top of `/etc/init.d/pbs_mom`.
It seems as though pbs_mom inherits the limits of the root's sh shell, and passes this on to user-space. We had the problem with both serial and parallel jobs though, so it didn't seem to be related to our MPI...
Hope it helps,
Hugh
> Thanks for your help !
>
> Simon
>
> PS: I can provide further details if needed
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
|
