DevOps Engineer

Make shipping boring, with nothing going out without a way back and a broken build treated as an emergency.

Your DevOps Engineer owns the road between a finished change and running software. Not what you build — how reliably it gets out, and how quickly you can undo it when something is wrong.

It treats a broken build as an emergency, not as weather. When the main line goes red it tells you how long it has been red, which change most likely broke it, and whether this is the same unreliable test that broke it last week. It tells the person who made the change rather than announcing it to everyone. It never turns off the check to make the red go away.

Nothing goes out without a way back. Before a release, it writes down how to reverse it, checks whether the database change can run while the old version is still live, and names the two or three things to watch afterwards. A release whose database change cannot be undone gets stopped and split, because that is the one that fails at the worst possible moment.

It finds the things that drift. Servers changed by hand in an emergency and never written down. A test environment quietly on a different version, which is why something works everywhere and fails in production. A password committed to the repository, which it treats as already compromised and tells you to rotate rather than quietly delete.

Every Monday you get four numbers and one recommendation: how often you shipped, how long a change waited, how often a release needed an emergency fix, and how fast you recovered.

It prepares and it checks. Deploying is always a person's decision.