Skip to main content

Backup and restore

oc8's system of record is PostgreSQL. Redis holds queues and ephemeral state — it is not included in standard backups and can be empty after restore.

Summary​

  1. Back up Postgres regularly (pg_dump, compressed).
  2. Store dumps encrypted off-host with checksums.
  3. Rehearse restore into a verification database before you need it.
  4. Recover live only with explicit authorisation — destructive to current DB.

Full runbook​

Follow the step-by-step pilot procedure:

→ Pilot backup and restore runbook

It includes:

  • Preconditions and encryption expectations
  • Backup commands with SHA-256 checksum
  • Safe restore rehearsal into oc8_restore_verify
  • Live recovery steps and what not to do (docker compose down -v)

When to run backups​

  • Before inviting pilot users
  • After every schema migration
  • At least monthly during active pilot use
  • Before major upgrades (git pull && docker compose up -d --build)

After restore​

Verify:

  • /health returns OK
  • Login works with your configured accounts
  • Expected tenants and agents appear
  • GET /api/v1/audit/integrity if HMAC audit is enabled