Skip to main content

Capacity Planning

Capacity planning focuses on object storage, sync frequency, backup retention, concurrent operations, and control-plane resources.

Object Storage Capacity​

Estimate:

Estimated capacity = protected PVC data size × retention points × change-rate factor + resource backup metadata

Factors:

  • Total PVC data size.
  • Whether backups are incremental.
  • Automatic backup schedule.
  • Retention policy.
  • Temporary data from drills and cross-cluster restores.

Production buckets need capacity alerts and lifecycle policies.

Sync Frequency​

Higher sync frequency lowers RPO but increases:

  • Source cluster IO.
  • Object storage writes.
  • Target cluster restore pressure.
  • Velero task count.
  • Operator reconcile pressure.

Use application tiers. Do not apply one high-frequency policy to all namespaces.

Concurrent Operations​

Concurrent failover, restore, or drill workflows compete for:

  • Target cluster scheduling resources.
  • Object storage throughput.
  • Velero controllers.
  • Network bandwidth.
  • Operator reconcile workers.

Use DisasterGroup level, parallelism, timeout, and retry policy to control concurrency.

Control Plane Resources​

Production guidance:

  • Run multiple Server replicas when needed.
  • Enable operator leader election.
  • Set requests/limits for operator, server, and web.
  • Send logs to centralized storage to avoid local log growth.

Validation​

Before production, run at least:

  1. Single-instance backup/restore load test.
  2. Single-instance failover drill.
  3. Multi-instance disaster group drill.
  4. Object storage capacity and throughput observation.
  5. Post-restore business health checks.

Test results should become business RPO/RTO baselines.