Slurm "run" is failing to find "munge" component when run with "container-name"

[RamHPC]

Slurm "run" is failing to find "munge" component when run with "container-name"

[rhc54]

Couple of possibilities:

1. the PMIx inside the container was not built with munge support
2. the container itself lacks the munge library

Slurm requires that the munge support be present, or else you will get those errors.

[RamHPC]

Yes. OpenMPI, PMIx and all other libraries are not part of the container. My expectation with Enroot/Pyxis is that it will utilize the installation from the host machine. Please suggest if that understanding is wrong. Slurm is working fine on the cluster which uses munge. I am not interested to build the container with all these packages like slurm, openmpi, pmix, ucx etc.

[rhc54]

I'm not sure I understand how you are intending the app inside the container to load all those libraries and their dependencies. What kind of container is this? Are you mounting the location where munge can be found, and is it mounted to the same path where PMIx expects to find it?

[RamHPC]

This looks like a problem in enroot or providing configurations to enroot. The PATH and LD_LIBRARY_PATH need to be exported to enroot for the container to find the correct libraries. Thank you for your earlier suggestion which provided a clue that the local "env" is not getting exported to enroot -> container. I am closing the issue here and follow up with Enroot/Pyxis project.