Closed holderg closed 4 years ago
slurm-sync-gcp only looks at machines matching the -compute* pattern. All other compute nodes will get terminated as un-responsive.
The correct filter is the intersection of the running instances and the node list from slurm.conf.
Fixes #11
Needed to rework slurm-sync-gcp filter after queuying for running instances.
verified arbitrary compute nodes are not being killed after 30 seconds, but idle nodes are still being taken down.
slurm-sync-gcp only looks at machines matching the-compute* pattern. All other compute nodes will get terminated as un-responsive.
The correct filter is the intersection of the running instances and the node list from slurm.conf.