What does a good CI pipeline look like for a backend service?

It runs on every push, returns feedback in under ten minutes, and blocks merge on failure. Order stages cheapest first: lint and type checks, unit tests, then integration tests against real dependencies in containers, then one build artifact you promote rather than rebuild per environment. Keep it deterministic, quarantine flaky tests instead of retrying blindly, and make the same checks runnable locally.

The first property is speed, because a pipeline that takes forty minutes is one people work around. I want lint and type checks failing inside a minute, unit tests in a few, and the whole thing under ten. So stages go cheapest first: static checks, unit tests, then integration tests against real Postgres and Redis in containers rather than mocks, because the bugs I care about live at the boundary. Then it builds one artifact, a container image tagged with the commit sha, and every environment after that deploys that exact image. Rebuilding per environment means the thing you tested is not the thing you shipped. On flaky tests I am strict: an automatic retry is a way of hiding a real race, so a flaky test gets quarantined and somebody owns fixing it that week. And I want the same commands runnable locally, usually behind a makefile, because debugging by pushing commits to see what CI says is a miserable way to spend an afternoon.