Backup and restore protects Kubernetes cluster state and application data against corruption, cluster loss, failed upgrades, and other recovery events. Effective protection separates control-plane state from persistent application data and validates the complete recovery path. section 5.2.1 section 3.2.2
Treat backup coverage as several operational responsibilities:
- Schedule Anthos admin-cluster backups in production and create on-demand backups before upgrading Anthos. section 5.2.1
- Back up user-cluster etcd when necessary, but recognize that this captures Kubernetes configuration rather than persistent volumes or application data. section 5.2.1
- Back up persistent volumes and other application-specific data separately to another storage device with a tool such as Velero. section 5.2.1
- Store etcd backups outside the cluster in a safe location, because etcd contains every Kubernetes resource, including sensitive secrets. section 3.2.2
- Practice restoring admin and user clusters together with application data to verify that the recovery process works. section 5.2.1
Recovering after an etcd failure
If an etcd cluster enters split-brain or becomes corrupted, recreate the etcd cluster from an external backup, then restore the cluster state. Persistent application data still requires its separate backup and recovery process. section 3.2.2 section 5.2.1
