Backups and Restore
VeloDB Cloud lets you create database-level export backups or warehouse-level lightweight snapshots, then restore data from a retained recovery point. Backups help you recover from accidental data changes, data corruption, and failed high-risk operations. They also provide recovery points for migrations and planned maintenance.
Backup and recovery are a shared responsibility. VeloDB Cloud provides the mechanisms to create, retain, monitor, and restore backup sets. You choose what to protect, how often backups run, how long they are retained, which compute resources Export Backup uses, and how applications return to service after recovery.
Backups do not replace high availability or cross-region disaster recovery. High availability keeps a workload available during infrastructure failures within a region. Backups provide retained recovery points, but they remain in the warehouse region. To protect against a regional outage, maintain a recoverable copy in another region. See Reliability and Disaster Recovery for how these controls work together.
Plan your backup and recovery strategy
Start with the business requirements for each workload:
- Recovery point objective (RPO): The maximum amount of recent data the workload can lose. Choose a backup schedule based on how frequently you need a completed recovery point. A scheduled run is not a usable recovery point until it completes successfully.
- Recovery time objective (RTO): The maximum acceptable time to restore service. Include backup selection, restore task startup, data restore or import, validation, and application cutover when estimating recovery time.
- Recovery scope: Identify whether you need to recover one database or the complete warehouse. Also record the SQL users, roles, connections, secrets, and other control-plane configuration that the workload needs.
- Retention: Retain enough successful backup sets to cover delayed discovery of data loss or corruption. Monitor expiration dates so that a required recovery point is not removed before it is tested or used.
Creating backups alone does not demonstrate that a workload can recover within its objectives. Review backup task status, investigate failed or skipped objects, and test restores regularly. Validate the restored schema, row counts, representative queries, permissions, connections, and application behavior before directing production traffic to the recovered warehouse.
Choose a backup mode
VeloDB Cloud provides two backup modes. Choose the one that matches the recovery scope you defined above, then follow its guide to create a plan.
- Export Backup: Backs up a selected database to object storage using a cluster you choose. Use it for database-level recovery, partial recovery, and migration. Export Backup is available in VeloDB Cloud 4.0 and later.
- Lightweight Snapshot: Backs up the entire warehouse from the underlying snapshot capability, with no object or cluster selection. Use it for fast full-warehouse recovery and recovery points before high-risk operations. Lightweight Snapshot is available in VeloDB Cloud 26.1 and later.
Note:
Lightweight Snapshot is in private preview. To use it, contact VeloDB Cloud Support to join the allowlist.
Compare backup modes
The following table summarizes how the two modes differ.
| Export Backup | Lightweight Snapshot | |
|---|---|---|
| Implementation | Built on the VeloDB Cloud export feature: exports data to object storage and imports it on restore. | Built on the underlying snapshot capability: copies only metadata references, not data. |
| Granularity | Database level. | Warehouse level. |
| Coverage | Limited to what the export feature supports. Objects with unsupported data types cannot be backed up. See Export Backup limitations. | The entire warehouse. |
| Main use cases | Logical backup, migration, and partial recovery. | Full warehouse recovery and recovery points before high-risk operations. |
| Backup and restore duration | Depends on data volume and the warehouse resources used for export and import. Larger datasets take longer. | Typically faster because the snapshot does not copy the data. Actual duration depends on the warehouse and snapshot workload. |
| Minimum version | VeloDB Cloud 4.0. | VeloDB Cloud 26.1. |
Choose Export Backup when
- You want to back up a specific database instead of the entire warehouse.
- You need to migrate a database to another warehouse.
- You want a relatively fine-grained periodic backup strategy.
Choose Lightweight Snapshot when
- You need to recover the entire warehouse after a warehouse-level failure or widespread data damage.
- You need a recovery point before a version upgrade or other high-risk operation.
- An operation may affect multiple databases, or you need to preserve warehouse-level objects such as users, permissions, and connections.
- You need to restore the whole warehouse with a low RTO.
- You want platform-managed warehouse-level recovery points.
Backup schedule
For either mode, choose how the backup runs when you create a plan:
- Periodic backup: Runs automatically on a regular schedule. Use it for recurring recovery points and regular operational protection.
- One-time backup: Runs once and does not create a recurring schedule. Use it before a specific operation, such as a schema change, data reload, or other planned maintenance.
Create a backup plan
To create an Export Backup plan, see Create an Export Backup plan.
To create a Lightweight Snapshot plan, see Create a Lightweight Snapshot plan.
Restore
After a backup set completes successfully, you can restore its data after accidental changes, data corruption, or a failed high-risk operation. Use Restore from a backup to restore an Export Backup to a selected database or a Lightweight Snapshot to an entire warehouse. Restore to a new warehouse when you need to validate the recovered data before switching production traffic.