Purpose
Every change to kanman is traceable, checked automatically before it is merged and checked in full before it reaches production.
How a change is made
- Every change lives in version control with a description of what it does and why.
- Before a change is merged, automated checks run on it: a scan for credentials in the commits, linting, type checks, the unit tests it affects, database tests when the schema changes, and end-to-end tests of the parts of the app it touches.
- Database changes are additive by default. A change that removes something in use is released only after the code that no longer needs it.
How a change reaches production
- Changes collect on an integration branch.
- Several times a weekday, the full test suite runs on that branch. Only a green run is promoted to production.
- An urgent fix takes the same path, started by hand; it never skips the tests.
- A failed promotion is fixed first, before anything else is merged.
Run machines and runners
- New versions of kanman’s run machines first take a quarter of the runs and are compared with the previous version on how those runs end. If the new version does worse, kanman goes back to the previous version automatically.
- When you update a self-hosted runner, its first runs are compared with the previous version in the same way, and we contact you when the new version does worse.
- Self-hosted editions update from signed releases; a failed update rolls back to the previous release’s files and database.
Review
This policy is reviewed when the release process changes, and at least once a year.