Hi Melissa, Probably best to post this question to [email protected] and go from there. Thanks, Ken On 05/01/2017 01:15 PM, Melissa Romanus wrote:

Hello,

I am experiencing issues on the SDSC Comet system when using the intel compilers with mvapich2. The scheduling system on Comet is slurm. It seems like the code is seg-faulting inside of MPI_Comm_dup, but prior to that, it seems like it is rejecting a connection request to "self" (i.e., same IP to same IP).

The modules loaded are:

``` $ module list

Currently Loaded Modulefiles: 1) intel/2013_sp1.2.144 2) mvapich2_ib/2.1 3) gnutools/2.69 ```

I am attempting to use the `ib0` interface. In my job script, I am launching 3 different applications. I am **not** using slurm `--multi-prog`. I am instead using 3 different `srun` commands. My job has to be launched this way.

Using OpenMPI, I can set the MCA parameters to allow connections from `self` at the byte-transfer layer, i.e., `OMPI_MCA_btl="self,openib"` and specify to slurm that I would like to use `--mpi=pmi2`.

I think the mvapich errors that I am experiencing stem from the fact that the "self" connection is rejected (i.e., node to itself). Is there a way to tell MVAPICH to allow the self connection? I think I want the `--with-device=ch3:nemesis:ib` command in some capacity, but I'm not sure if that would be enough to allow the connection from the node to itself. Is the self connection inherently a TCP connection? Or do I still need `--mpi=pmi2` for srun? Can I use srun or do I need to use `mpiexec` explicitly?

Could this also be the cause of the error described by this FAQ? https://wiki.mpich.org/mpich/index.php/Frequently_Asked_Questions#Q:_All_my_....

Any help you can provide is greatly appreciated.

-Melissa

_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss