Open MPI logo

Open MPI User's Mailing List Archives

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

Subject: [OMPI users] experiencing hangs with inter-communicator collectives
From: sreeram potluri (potluri_at_[hidden])
Date: 2009-07-15 16:32:17


Hi OpenMPI Team,

I am trying to run a simple application that does an alltoall over an
intercommunicator and I experience hangs when I run more than 3 processes
per node. A similar program that uses intra-communicator completes fine with
upto 8 processes per node.

This is the error message I see in case of inter-communicator:

[node20-oib][[61039,1],2][btl_openib.c:201:adjust_cq] cannot resize
completion queue, error: 22
[node20-oib][[61039,1],3][btl_openib.c:201:adjust_cq] cannot resize
completion queue, error: 22
[node20-oib][[61039,1],0][btl_openib.c:201:adjust_cq] cannot resize
completion queue, error: 22
[node21-oib][[61039,1],7][btl_openib.c:201:adjust_cq] cannot resize
completion queue, error: 22

I am using OpenMPI 1.3.3

This is how I run the applicaiton: ../bin/mpiexec --prefix
/home/potluri/openmpi-gcc/ --hostfile hostfile2 -np 8 ./alltoall_inter

The nodes i am using are Intel Clovertowns with 8 cores and they have
ConnectX cards.

Any insights into why this could be happening will be very helpful.

Thanks
Sreeram Potluri