Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

OpenMPI/2.1.1-GCC-6.4.0-2.28 build fails if ucx-devel is installed on system #5805

Closed
jabl opened this issue Feb 9, 2018 · 8 comments
Closed
Milestone

Comments

@jabl
Copy link

jabl commented Feb 9, 2018

The easyconfig OpenMPI/2.1.1-GCC-6.4.0-2.28 (part of the foss/2017b toolchain) fails to build with the error below:


Making all in mca/pml/ucx
make[2]: Entering directory `/l/easybuild/build/OpenMPI/2.1.1/GCC-6.4.0-2.28/openmpi-2.1.1/ompi/mca/pml/ucx'
  CC       pml_ucx.lo
  CC       pml_ucx_request.lo
  CC       pml_ucx_datatype.lo
  CC       pml_ucx_component.lo
  CCLD     libmca_pml_ucx.la
libtool:   error: require no space between '-L' and '-lrt'
make[2]: *** [libmca_pml_ucx.la] Error 1

Removing the rpm package "ucx-devel" fixes it. ucx-devel is installed as part of Mellanox OFED drivers (this is on CentOS 7). Most likely the culprit is open-mpi/ompi#4345

@boegel
Copy link
Member

boegel commented Feb 11, 2018

@jabl Thanks for reporting this...

Since this is a bug upstream, how do you suggest we handle this, try to patch the configure script?

@boegel boegel added this to the next release (3.5.2 or 3.6.0) milestone Feb 11, 2018
@jabl
Copy link
Author

jabl commented Feb 23, 2018

Yeah, well, seems 2.1.2 doesn't have the fix and 2.1.3 isn't out yet, so yeah, I guess adding ggouaillardet/ompi@7a3ace1 to the easyconfig file could do the trick (I haven't tested that myself, I just worked around it by temporarily uninstalling ucx-devel)

@boegel
Copy link
Member

boegel commented Feb 23, 2018

@wpoely86 This is the same issue you were seeing in #5768?

@wpoely86
Copy link
Member

Yes, it's the same.

@boegel boegel modified the milestones: 3.5.2, 3.x Feb 24, 2018
@boegel
Copy link
Member

boegel commented Feb 25, 2018

@jabl It's not as simple as just applying the patch from ggouaillardet/ompi@7a3ace1, we'll need to regenerate the configure script too...

@wpoely86 Do you think it's worth the trouble to try and patch this? I'm inclined to try and fix this since this affects OpenMPI-2.1.2-GCC-6.4.0-2.28.eb which is part of foss/2018a...

Can we come up with a direct patch for the configure script?

@boegel boegel modified the milestones: 3.x, 3.5.2, 3.6.0 Feb 25, 2018
@boegel
Copy link
Member

boegel commented Mar 5, 2018

Or maybe we should include --without-ucx in the easyconfig for this particular Open MPI version to dance around this, as suggested by @hajgato on the EasyBuild mailing list?

@boegel
Copy link
Member

boegel commented Mar 5, 2018

@jabl @wpoely86 Proposed @OleHolmNielsen Proposed workaround in #5949

@boegel
Copy link
Member

boegel commented Mar 6, 2018

follow-up in #5949

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants