FIELD NOTES / PECKING PRACTICES

Writing Good Runbooks at 3 a.m.

How to write operational docs that help a sleepy person at the worst time.

August 18, 20266 min readBy Gertrude Gizzard
In shortStart with symptoms / Number the steps / Link the dashboards / Test the runbook.

1. Start with symptoms

People search by what they see, not by what it means. Title it "Checkout fails with 504", not "Gateway handling".

2. Number the steps

Each step should be one action and one expected result. Short sentences. No cleverness.

3. Link the dashboards

The right graph, pre-filtered to the right time, saves 10 minutes.

4. Test the runbook

Hand it to someone who does not know the system. If they get stuck, you have found a bug.

Written by Gertrude Gizzard, Head of People. Disagree? Tell us. We read every message, usually before the second coffee.