Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: Re: [OMPI users] modified hostfile does not work with openmpi1.7rc8
From: Gustavo Correa (gus_at_[hidden])
Date: 2013-03-17 22:55:33

Hi Tetsuya Mishima

In your example, have you tried not to modify the node file,
launch two mpi processes with mpiexec, and request a "-bynode" distribution of processes:

mpiexec -bynode -np 2 ./my_program

This should launch one MPI process in each of two nodes.
See 'man mpiexec' for details.
Works in OMPI 1.6.
I don't know about OMPI 1.7.

You can continue to ask for two nodes with four processors/cores each in your Torque script,
#PBS -l nodes=2:ppn=4
to use the remaining cores for the OpenMP threads.
I hope this helps,
Gus Correa

On Mar 17, 2013, at 8:52 PM, tmishima_at_[hidden] wrote:

> Hi openmpi developers,
> I have been evaluating our FEM aplication with new openmpi-1.7rc7 under
> Torque job controler.
> Now I encountered a trouble that "-hostfile" does not work properly.
> Since my application is hybrid(MPI+OpenMP), I have to modify
> $PBS_NODEFILE and use "-hostfile".
> I don't add new hosts to the hostfile according to FAQ. It's just a
> subset of the hosts allocated to the Torque. At leaset, this method
> works well with openmpi-1.6.x.
> I hope this issue will be fixed in the next release of openmpi-1.7.
> Best Regards,
> Tetsuya Mishima
> (1) Example of 2MPI having 4 threads:
> $PBS_NODEFILE -> modified hostfile
> node01 node01
> node01 node02
> node01
> node01
> node02
> node02
> node02
> node02
> (2) The error message I got is as follows:
> --------------------------------------------------------------------------
> A hostfile was provided that contains at least one node not
> present in the allocation:
> hostfile: pbs_hosts
> node: node01
> If you are operating in a resource-managed environment, then only
> nodes that are in the allocation can be used in the hostfile. You
> may find relative node syntax to be a useful alternative to
> specifying absolute node names see the orte_hosts man page for
> further information.
> --------------------------------------------------------------------------
> _______________________________________________
> users mailing list
> users_at_[hidden]