Runbook
A runbook is a documented procedure for a specific operational task or alert. It tells the responder how to check what is wrong, how to fix or mitigate it, and when to escalate. A good runbook lets any engineer on the on-call rotation handle a problem without the one person who knows the system. It is also called a playbook.
How it works
Each alert links to its runbook. A runbook for "Postgres disk above 90%" might contain:
What this alert means and the user impact.
Checks: run
df -hand look for large WAL or temporary files.Mitigation: remove files safely, or grow the volume by 50 GB.
Escalation: page the database owner if usage passes 95%.
After each incident, update the runbook with what was missing. Steps that run the same way every time are candidates for automation, which reduces toil.
Why it matters in production
Runbooks cut mean time to recovery because responders stop guessing at 2 a.m. They spread knowledge beyond a few experts. A stale runbook is dangerous, so review them like code.
See also
On the One2N blog
Further reading
