Skip to main content

Backup & Disaster Recovery

What to back up

DataCadenceRetention
Postgres (Farm, Activity, IAM)continuous (WAL) + nightly snapshot30 days hot, 1 year cold
Object store (rasters, AI outputs)versioning enabledper project policy
Sensor observations (time-series store)nightly snapshot30 days hot, 5 years cold
Keycloak realm configon changeindefinite
Audit logcontinuous, append-onlyregulatory minimum

Restore drill

Run a full restore-from-cold quarterly. The drill must include:

  • Bring up a fresh stack against the restored data
  • Verify a representative service run end-to-end
  • Verify the audit log is intact

DR scenarios

  • Single-service failure — Kubernetes restart, automatic.
  • Regional outage — failover to a warm standby region (if configured).
  • Data corruption — roll back to the last clean snapshot; replay the audit log to identify affected commissions.