[mpich-discuss] Not able to run MPI program parallely...

[Pavan Balaji]

On 05/01/2012 01:39 PM, Albert Spade wrote:
> [mpiexec at beowulf.master] Launch arguments:
> /opt/mpich2-1.4.1p1/bin/hydra_pmi_proxy --control-port
> beowulf.master:60190 --debug --rmk user --launcher ssh --demux poll
> --pgid 0 --retries 10 --proxy-id 0
> [mpiexec at beowulf.master] Launch arguments:
> /opt/mpich2-1.4.1p1/bin/hydra_pmi_proxy --control-port
> beowulf.master:60190 --debug --rmk user --launcher ssh --demux poll
> --pgid 0 --retries 10 --proxy-id 1
> [mpiexec at beowulf.master] Launch arguments:
> /opt/mpich2-1.4.1p1/bin/hydra_pmi_proxy --control-port
> beowulf.master:60190 --debug --rmk user --launcher ssh --demux poll
> --pgid 0 --retries 10 --proxy-id 2
> [mpiexec at beowulf.master] Launch arguments:
> /opt/mpich2-1.4.1p1/bin/hydra_pmi_proxy --control-port
> beowulf.master:60190 --debug --rmk user --launcher ssh --demux poll
> --pgid 0 --retries 10 --proxy-id 3

Thanks.  It looks like mpiexec thinks that all four nodes are local 
nodes, which is a bug that was fixed after 1.4.1p1 was released.  Can 
you try the mpiexec from 1.5b1:

http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/1.5b1/hydra-1.5b1.tar.gz

You don't need to recompile the application.  Just run it with this new 
mpiexec.

  -- Pavan

-- 
Pavan Balaji
http://www.mcs.anl.gov/~balaji