Back up and restore an installation

Prepare a recovery procedure before accepting production traffic. You need administrative access to the data services and an isolated environment for the restore test.

Define the recovery set

Back up PostgreSQL with a database-aware method, copy the audio object store using its supported backup mechanism, and retain the deployment configuration and required secrets in a separate protected location. Include Redis persistence when queue and campaign recovery depends on it. Record the application image tag and database migration version with each recovery set.

Keep records and objects consistent

Choose a maintenance window or a coordinated snapshot method. A copied running database volume alone is not evidence of a recoverable database. Verify that the restored conversation records can access the matching recording and transcript objects.

Rehearse recovery

Restore into an isolated installation using the recorded release. Keep outbound calls, campaign workers and external webhooks from contacting real customers during the rehearsal. Check login, a known conversation record, a stored artifact and a controlled browser call. Document the elapsed recovery time and any missing data before approving the procedure for production use.