On Wed, 17 Dec 2003, Per Madsen wrote:
> I’m trying to run a parallel application with mpirun and lam 6.5.9.
> and I’m getting the following error messages:
>
> mpirun (set_stdio): Too many open files in system
>
> The application I’m running is a little special as it consists of a
> parallel main task for a iterative solver for solving a large system of
> equations (dimension up to ~300.000.000). After some iterations in the
> main task, the master processor make a system call in order to start a
> parallel subtask by mpirun on the same nodes as the main task The
> subtask manipulates the data used in the main task. When the subtask has
> completed, the main task is resumed. When I run this application on 16
> nodes, the error occurs after 18 successful invocation of the subtask.
We fixed a file descriptor leak in the lamd in version 7.0 that *may* be
the source of your problem.
Is there any chance that you could download the latest (7.0.3) and give
your application a whirl with it?
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
