MPI seems broken in the IntelPython docker images

[intelpython]

MPI seems broken in the IntelPython docker images

[TylerBackman]

In native anaconda on OSX, I do not have this error- mpi4py works properly.

[rscohn2]

Thanks for reporting it. I will check it on Monday.

Robert

[rscohn2]

Hi Tyler,

I am not able to reproduce the problem. Your message above does not show the 'Python 3.6.3 ..' message. Is that something you filtered out? This is docker, so I would only expect the underlying kernel and version of docker to matter. I included that information below. What do you see?

rscohn1@fxsatlin01:~/local/ws$ docker run -it --rm intelpython/intelpython3_full:2018.0.1 python
Python 3.6.3 |Intel Corporation| (default, Oct 16 2017, 15:28:36)
[GCC 4.8.2 20140120 (Red Hat 4.8.2-15)] on linux
Type "help", "copyright", "credits" or "license" for more information.
Intel(R) Distribution for Python is brought to you by Intel Corporation.
Please check out: https://software.intel.com/en-us/python-distribution
>>> from mpi4py import MPI
>>>

rscohn1@fxsatlin01:~/local/ws$ uname -a
Linux fxsatlin01 4.4.0-116-generic #140-Ubuntu SMP Mon Feb 12 21:23:04 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux
rscohn1@fxsatlin01:~/local/ws$ docker version
Client:
 Version:      1.12.6
 API version:  1.24
 Go version:   go1.6.4
 Git commit:   78d1802
 Built:        Tue Jan 10 20:38:45 2017
 OS/Arch:      linux/amd64

Server:
 Version:      1.12.6
 API version:  1.24
 Go version:   go1.6.4
 Git commit:   78d1802
 Built:        Tue Jan 10 20:38:45 2017
 OS/Arch:      linux/amd64
rscohn1@fxsatlin01:~/local/ws$

[TylerBackman]

I am running docker on OSX so I ran uname within the container so you can see the kernel version. See below for the full output:

$ docker run -it --rm intelpython/intelpython3_full:2018.0.1 /bin/bash
root@a8a8910fd98f:/# python
Python 3.6.3 |Intel Corporation| (default, Oct 16 2017, 15:28:36)
[GCC 4.8.2 20140120 (Red Hat 4.8.2-15)] on linux
Type "help", "copyright", "credits" or "license" for more information.
Intel(R) Distribution for Python is brought to you by Intel Corporation.
Please check out: https://software.intel.com/en-us/python-distribution
>>> from mpi4py import MPI
Segmentation fault
root@a8a8910fd98f:/# uname -a
Linux a8a8910fd98f 4.9.36-moby #1 SMP Wed Jul 12 15:29:07 UTC 2017 x86_64 GNU/Linux

Docker version output running on OSX:

$ docker version
Client:
 Version:      17.06.0-ce
 API version:  1.30
 Go version:   go1.8.3
 Git commit:   02c1d87
 Built:        Fri Jun 23 21:31:53 2017
 OS/Arch:      darwin/amd64

Server:
 Version:      17.06.0-ce
 API version:  1.30 (minimum version 1.12)
 Go version:   go1.8.3
 Git commit:   02c1d87
 Built:        Fri Jun 23 21:51:55 2017
 OS/Arch:      linux/amd64
 Experimental: true

[TylerBackman]

Note also that I was able to get mpi running for test purposes by building my own Docker container on top of your container, but compiling mpich from source, and then adding the mpi4py from pip instead of the Intel one.

The reason I want to use mpi in your container, is because we're using it for a supercomputing shifter base image. Here is an example of an MPI docker image built on Ubuntu instead of Intel Python:
https://github.com/NERSC/base-images/blob/master/ubuntu-mpi/Dockerfile

[rscohn2]

Someone else on the team has reproduced the problem on a mac and we are investigating.