stackhpc / ansible-slurm-appliance

A Slurm-based HPC workload management environment, driven by Ansible.
43 stars 18 forks source link

Detection of block device prefix in CaaS failed #343

Closed sjpb closed 9 months ago

sjpb commented 10 months ago

From @mkjpryor:

issue with the Slurm appliance turned out to be a problem with community images and the detection of the block device prefix. Specifically, it appears that the openstack.cloud.image_info module does not consider community images by default, instead returning an empty image list. I fixed it temporarily at Cambridge by specifying the block device prefix explicitly. ... getting the duplicate images error [from here] (which actually is a "not exactly one" error, as in zero images would also trigger it)

My testing showed that the above code does return the community image. There might be an os collection version problem (this was pre-merge CaaS slurm, but the failing code will have been merged).

It occcured when trying to use community images uploaded in the -cloud-portal-prod tenancy to deploy a cluster in the -cloud-portal-demo tenancy, so it may also be a cross-project image problem.

I think #341 should fix this, as the need to get properties from images will go away.

sjpb commented 9 months ago

Fixed by #341