With any major release of slurm-gcp
, there has been architectural changes to
terraform modules which impact backwards compatibility.
Please review CHANGELOG for documented changes.
_slurm_metadata_devel
_slurm_destroy_subscriptions
_slurm_notify_cluster
Upgrading will be destructive and any running workload will be terminated and lost.
The following steps assume the slurm_cluster module is managing the cluster deployment.
- Fully drain the v5 cluster.
terraform destroy
the v5 cluster.- Upgrade source from v5 to v6.
- Build new images for v6 cluster.
- Update v6 tfvars based on v5 tfvars.
terraform apply
the v6 cluster.