With parallel backup processing in OpenShift api for Data Protection, you can back up several applications from different namespaces simultaneously. This concurrent processing ensures that a single large backup task does not delay the backups for other applications in your cluster.
By default, Velero processes backup resources sequentially - one resource in the InProgress phase at a time. Parallel backup changes this behavior to handle backups concurrently.
The following considerations apply for enabling parallel backups in OADP:
Resource allocation: Increasing the concurrency settings requires additional CPU and memory for the Velero container. Monitor resource consumption to ensure cluster stability.
Backup scope: Parallel processing provides the most benefit for backups that contain a large number of Kubernetes resources or many smaller volumes. Backups dominated by a few large volumes see less benefit because most time is spent waiting for asynchronous data movement to complete.
Namespace partitioning: To keep data integrity, OADP uses namespace-level isolation when processing parallel backups. Two backups cannot run simultaneously if they share any included namespaces.
| Scenario | Included namespaces | Processing behavior |
|---|---|---|
Backup A and Backup B |
|
Parallel: The namespaces are isolated. |
Backup A and Backup B |
|
Serialized: The shared namespace creates a lock. |
Full cluster and Backup A |
All namespaces and |
Serialized: A cluster-wide backup includes all namespaces and locks the entire cluster. |