← All runbooks
Under construction · working draft requires a release manifest
Technical runbook
Upgrade and rollback
Controlled version change without relying on the latest tag and without irreversible retry in production.
01
Before the change
- Verify the supported path from the current version and the compatibility of all enabled services.
- Review breaking changes, configuration, schema, API, and template changes.
- Back up your data and perform or document a current recovery test.
- Download new images before the maintenance window and record digests.
- Establish the last safe moment to return and the person responsible for the go/no-go decision.
02
Execution
- Stop new inputs or switch the system to runbook mode.
- Verify that the integration queues are empty or safely preserved.
- Run the migrations in the prescribed order and monitor their results.
- Launch services with new digests and perform technical and process smoke tests.
- Only resume full operation after acceptance.
03
Rollback is not just an old image
If the upgrade changed the data or schema, rolling back the container alone is not enough.
- Use only the described reverse migration, or restore a consistent backup.
- Return a compatible set of all services together.
- Consider the documents and messages created after the backup.
- After returning, repeat the smoke test and reconciliation of integration queues.
- Record the incident and the decision to return in the change log.
Next runbook