mpi4py仅在mpiexec下工作

Question

I have set up mpi4py on a new server, and it isn't quite working. When I import mpi4py.MPI , it crashes. However, if I do the same thing under mpiexec , it works. On my other server and on my workstation, both techniques work fine. What am I missing on the new server?

Here's what happens on the new server:

$ python -c 'from mpi4py import MPI; print("OK")'
--------------------------------------------------------------------------
It looks like orte_init failed for some reason; your parallel process is
likely to abort.  There are many reasons that a parallel process can
fail during orte_init; some of which are due to configuration or
environment problems.  This failure appears to be an internal failure;
here's some additional information (which may only be relevant to an
Open MPI developer):

  PMI2_Job_GetId failed failed
  --> Returned value (null) (14) instead of ORTE_SUCCESS
--------------------------------------------------------------------------
--------------------------------------------------------------------------
It looks like orte_init failed for some reason; your parallel process is
likely to abort.  There are many reasons that a parallel process can
fail during orte_init; some of which are due to configuration or
environment problems.  This failure appears to be an internal failure;
here's some additional information (which may only be relevant to an
Open MPI developer):

  orte_ess_init failed
  --> Returned value (null) (14) instead of ORTE_SUCCESS
--------------------------------------------------------------------------
--------------------------------------------------------------------------
It looks like MPI_INIT failed for some reason; your parallel process is
likely to abort.  There are many reasons that a parallel process can
fail during MPI_INIT; some of which are due to configuration or environment
problems.  This failure appears to be an internal failure; here's some
additional information (which may only be relevant to an Open MPI
developer):

  ompi_mpi_init: ompi_rte_init failed
  --> Returned "(null)" (14) instead of "Success" (0)
--------------------------------------------------------------------------
*** An error occurred in MPI_Init_thread
*** on a NULL communicator
*** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
***    and potentially your MPI job)
[Octomore:45430] Local abort before MPI_INIT completed successfully; not able to aggregate error messages, and not able to guarantee that all other processes were killed!

If I run it with mpiexec , it's fine.

$ mpiexec -np 1 python -c 'from mpi4py import MPI; print("OK")'
OK

I'm running on CentOS 6.7. I've installed Python 2.7 as a software collection, and I've loaded the openmpi/gnu/1.10.2 module. MPICH and MPICH2 are also installed, so they may be conflicting with OpenMPI. I haven't loaded the MPICH modules, though. I'm running Python in a virtualenv:

$ pip list
mpi4py (2.0.0)
pip (8.1.2)
setuptools (18.0.1)
wheel (0.24.0)

Answer 1

It turned out that mpi4py is not compatible with version 1.10.2 of OpenMPI. It works fine with version 1.6.5.

$ module load openmpi/gnu/1.6.5
$ python -c 'from mpi4py import MPI; print("OK")'
OK