Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] Openmpi-1.5.3 issue " initialization failure on /dev/ipath (err=23)"
From: Ralph Castain (rhc_at_[hidden])
Date: 2012-03-29 10:59:38


One thing stands out right away: why are you specifying a hostfile? Did you remember to configure OMPI with --with-tm so we launch via Torque? If not, then you could hit issues as you are actually attempting to launch via ssh, which has implications on a Torque-based system.

On Mar 29, 2012, at 8:51 AM, Raju wrote:

> Hi Team,
>
> I am using Qlogic Infiniband and Openmpi-1.5.3. I can able to run the jobs by CLI without any issues, but when iam submitting over torque scheduler facing the below issue.
>
> I am facing issue while submitting the jobs through Torque scheduler. Error file is attached
>
> Overview of the problem:
>
> node1.ibab.ac.in.5910Driver initialization failure on /dev/ipath (err=23)
> --------------------------------------------------------------------------
> PSM was unable to open an endpoint. Please make sure that the network link is
> active on the node and the hardware is functioning.
>
> Error: Failure in initializing endpoint
>
> I gone through the link http://www.open-mpi.org/community/lists/users/2011/12/17888.php for solution, same followed but no luck.
>
> I exported the value in my input submit script file as export PSM_SHAREDCONTEXTS_MAX=16, and submitted the job.
>
> Sample inputfile is
>
> #!/bin/bash
> #PBS -N matmul
> #PBS -l nodes=1:ppn=1
> node=1
> ppn=1
> nprocs=`expr ${node} \* ${ppn}`
> echo "--- PBS_NODEFILE CONTENT ---"
> cat $PBS_NODEFILE
> export PSM_SHAREDCONTEXTS_MAX=16
>
> mpirun -np ${nprocs} --hostfile $PBS_NODEFILE /home/khan/a.out < /home/khan/iter
>
> Please let me know I doing correct or not ? and suggest me for best out ?
>
> Regards,
> Bhagya Raju K
> <errfile.txt>_______________________________________________
> devel mailing list
> devel_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/devel