Tagging disks to avoid garbage collection
During backup, restore, upgrade, or suspend operations on CML workspaces, EBS or Azure disks can be left in a temporarily detached or unmanaged state when the associated EKS or AKS cluster is still running. If there is a garbage collection script running, and it is not properly configured, the disks used by the workspace can be deleted unintentionally. If the disk is not backed up, then the data will be lost.
Garbage collection scripts need to check for the following tags and ignore disks that are tagged with the corresponding values.
Cloud | Tag key | Tag value |
---|---|---|
Azure | k8s-azure-created-by |
kubernetes-azure-dd |
AWS | kubernetes.io/cluster/liftie-* |
owned |