Outdated openmpi in xp65 analysis3

Technical
1

Hi all.

I am trying to use a package that requires openmpi in the analysis3 conda environment. However, it appears as if this is still an old install from hh5 which has been deprecated. Is there a way to update this to an install on xp65?

[mb0427@gadi-login-06 ~]$ ompi_info Package: Open MPI hh5_apps@gadi-login-05.gadi.nci.org.au Distribution Open MPI: 4.1.6 Open MPI repo revision: v4.1.6 Open MPI release date: Sep 30, 2023 Open RTE: 4.1.6 Open RTE repo revision: v4.1.6 Open RTE release date: Sep 30, 2023 OPAL: 4.1.6 OPAL repo revision: v4.1.6 OPAL release date: Sep 30, 2023 MPI API: 3.1.0 Ident string: 4.1.6 Prefix: /g/data/hh5/public/apps/openmpi/4.1.6 Configured architecture: x86_64-conda-linux-gnu Configure host: gadi-login-05.gadi.nci.org.au Configured by: hh5_apps Configured on: Thu Jun 6 01:16:20 UTC 2024 Configure host: gadi-login-05.gadi.nci.org.au Configure command line: ‘–prefix=/g/data/hh5/public/apps/openmpi/4.1.6’

Xarray: libmpi.so.40 error
2

Hi @Michael_barnes,

The ompi_info gives you the details of how openmpi has been compiled. I moved the installation to xp65 without recompiling the code to make sure things were back compatible after the move to xp65.
Any reason you need openmpi to be updated? 4.1.6 is the last version of the 4 series so there is not much reason to update. We are working on making things compatible with openmpi 5.0

Romain

3

So just to clarify, the openmpi install is not outdated. Some info from ompi_info are.

5

Thanks @rbeucher. It’s not that the version is outdated for my application, it is that there seems to be some connection to hh5. For example, error messages are not able to be created:
e.g.
mpirun -np 1 wrong_name--------------------------------------------------------------------------Sorry! You were supposed to get help about: orterun:exe-not-foundBut I couldn’t open the help file: /g/data/hh5/public/apps/openmpi/4.1.6/share/openmpi/help-orterun.txt: No such file or directory. Sorry!--------------------------------------------------------------------------

6

Thanks Let me have a look.

7

OK, I think I will need to recompile openmpi then…Those paths are hardcoded into the executable and there is no easy way around. I’ll do that ASAP. Sorry for the inconvenience

8

Thanks @rbeucher! I appreciate your help!

9

Hi @Michael_Barnes ,

Can you give it a try? I think I fixed it.

10

Thanks @rbeucher - on first inspection it looks great! See my little test below. I will continue to test and let you know if I find anything.

Singularity> mpirun -np 1 hostname
gadi-cpu-bdw-0006.gadi.nci.org.au
Singularity> mpirun -np 1 Random_name
--------------------------------------------------------------------------
mpirun was unable to find the specified executable file, and therefore
did not launch the job.  This error was first reported for process
rank 0; it may have occurred for other processes as well.

NOTE: A common cause for this error is misspelling a mpirun command
      line parameter option (remember that mpirun interprets the first
      unrecognized command line token as the executable).

Node:       gadi-cpu-bdw-0006
Executable: Random_name
--------------------------------------------------------------------------
11

Good. Let me know if you find anything wrong