CI/CD Pipeline Checks for Repeatable Releases

A useful deployment pipeline produces repeatable evidence that a particular version can be released. Review the path from source code to production, including failures and recovery.

Separate caches from release artifacts

A dependency cache speeds up a build; it is not the record of what was deployed. Use cache keys that reflect relevant dependencies and tooling. Preserve identifiable release artifacts and avoid placing credentials or other secrets in caches.

Make required checks explicit

Connect deployment jobs to the tests and checks they require. Document any emergency path and its follow-up checks. Treat flaky tests as defects to investigate rather than silently ignoring their results.

Control deployment access

Use environments, restricted credentials and approval rules appropriate to the project. Prevent overlapping production deployments when they would interfere with one another.

Rehearse recovery

Keep the previous known-good artifact available and test redeployment. Database changes may require a separate recovery plan: reverting application code alone may not reverse data changes.

Expose the evidence

Link the deployed version to its build, test results and release notes. Agree on who responds to a failed release and what signals trigger rollback.

Sources and further reading