On Fri, 23 May 2003, Yu Chen wrote:
> > What happens when you try to manually run the collector, even on one node?
> > Can you send the output? Can you also try running:
> >
> > cd /lamtest-6.5.9/reporting
> > ./collector
>
> %cd /lamtest-6.5.9/reporting
> %mpirun -s h n0 collector
> mpirun: cannot start collector on n0 (o): No such file or directory
> [snipped]
> So I guess something to do with the "-s h" switch, what does that do?
Appologies in the delay in replying; we're trying to get LAM/MPI 7.0 out
the door. :-)
"-s h" means "send from here". i.e., send the executable from the node
that you are running on and then go run it on n0. So for some reason,
it's either failing when it sends the executable from h to n0, or the file
becomes unavailable after it's received by n0.
Is n0 the same as h, in this case (i.e., is the node that you are running
on the same as the first node in your hostfile)? You can run the lamnodes
command to see which node is n0.
Can you try the latest 7.0 beta (www.lam-mpi.org/beta/) and see if the
same problem occurs?
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
