Closed marcodelapierre closed 1 year ago
The main commands that need to be run are
-B /var/run/munge/munge.socket.2,/etc/slurm
for a non-configless slurm deployment. I mount munge
to ensure authentication. The configuration for slurm is in /etc/slurm
.
Note that the versions of within the container likely have to be VERY close to that running on the host for this work. Currently I have tried where the differences are at the patch level and there are warnings so there could be unforeseen behaviour. Larger differences can result in errors if using plugins.
Thanks Pascal!
Would this image be able to run standalone, e.g. on a VM, and act as a Slurm simulator? Or is it only supposed to be a base image to build containers for Setonix?
Only in the former case, it would be very good to contribute a corresponding SHPC manifest here: https://github.com/singularityhub/shpc-registry/tree/main/quay.io/pawsey
https://github.com/PawseySC/pawsey-containers/blob/82c412e878793b804e1e178b2c8a0e2b4935ce0a/slurm/slurm-base/buildslurm.dockerfile#L103
@pelahi can you comment on which bind mounts and/or Singularity configs and/or host configs required for this container to work. Also, what is the scope: Setonix-only, or any machine? Thank you