I am a little confused on data backups and recovery in shared data mode. The documentation for backups and restores says: > Shared-data StarRocks clusters do not support data BACKUP and RESTORE. But then what other options do I have for moving data to another cluster, etc. If you are using the Celerdata service (with bring your own cloud) how do you configure it to load with a cluster snapshot at startup? There does not seem to be a place to do that at all in the Celerdata UI. In the documentation its clearly a whole YAML spec and disaster setting you turn on for the kubernetes operator. See: https://docs.starrocks.io/docs/4.0/administration/cluster_snapshot/

In CelerData Cloud BYOC, the Cluster Snapshot feature (introduced in StarRocks v3.4.2 for shared-data clusters) is handled differently than in a self-managed environment. Because CelerData manages the underlying infrastructure and configuration files, you do not have direct access to

fe/conf/cluster_snapshot.yaml

or the ability to manually restart FE nodes with the

--cluster_snapshot

flag. Instead, the restore process is integrated into the CelerData Cloud Console. Here is how you typically handle this: 1. Using the Cloud Console UI For BYOC clusters, "Cluster Snapshot" is primarily an automated disaster recovery mechanism. If you need to restore a cluster or recover metadata: * Navigate to the Backups/Snapshots Tab: In the CelerData Cloud Web UI, go to your cluster details and look for the Backups or Snapshots section. * Restore to a New Cluster: The standard workflow for a full cluster restore from a snapshot in CelerData Cloud is to create a new cluster and select "Restore from Backup" as the creation method. You will be prompted to select the snapshot source (your S3 bucket) and the specific snapshot/timestamp. * Metadata Recovery: If you are trying to recover from a total FE failure where the metadata is lost but the data remains in S3, CelerData's management plane usually handles the FE recovery automatically. If you need to trigger a manual restore from a specific URI that isn't appearing in the UI list, you may need to use the CelerData Cloud API or CelerData CLI. 2. Why you don't see the YAML spec The YAML-based method described in the StarRocks documentation is designed for self-managed deployments (where you manage the VMs and process lifecycles). In CelerData Cloud: * The Control Plane generates the necessary configuration and passes it to the nodes during the "Restore" or "Rebuild" task. * The UI abstracts the

cluster_snapshot.yaml

and the

--cluster_snapshot

startup flag. 3. If you have a custom Snapshot URI in S3 If you have a snapshot stored at a custom URI (e.g., from a different cluster or region) and want to use it for a "Cross-Cluster Recovery" in BYOC: 1. Ensure the Storage Volume or Backup Repository is correctly registered in the new CelerData environment. 2. During the Create Cluster process in the Web UI, choose the restore option. 3. If the specific snapshot does not appear, check the permissions of the IAM role attached to the CelerData BYOC environment to ensure it has

s3:Get_

and

s3:List_

access to that specific bucket/path. Recommendation If you are performing a critical production recovery and the UI does not provide the granularity you need (like pointing to a specific URI for

cluster_snapshot_path

), it is recommended to contact CelerData Support. In BYOC environments, certain low-level metadata recovery operations are performed via the backend management plane to ensure the cluster remains in a "Managed" state. Key difference to remember: * Self-Managed: You edit

cluster_snapshot.yaml

and run

start_fe.sh --cluster_snapshot

. * CelerData Cloud BYOC: You use the "Restore" action in the Cloud Console or the CelerData Cloud API, which executes those same steps on your behalf. References • administration/cluster_snapshot.md • administration/management/Backup_and_restore.md