liveBook · Manning

Runbook

As explained in Machine Learning System Design

A runbook is an operational document containing procedures for diagnosing and resolving recurring system problems.

A runbook is an operational document that helps on-call engineers diagnose recurring production problems, apply known fixes, and escalate issues when necessary. section 16.1

A useful runbook records:

  • Procedures for common failures and recurring load patterns. section 16.1
  • The people or vendors to contact when the on-call team cannot resolve an incident. section 16.1
  • The tools needed for investigation, including logs, metrics, alerts, dashboards, configuration, and source documentation. section 16.1

Handling a usage spike

When a large customer creates a sudden load increase, the runbook can provide a recipe for starting additional system instances and scaling them down after demand returns to normal. section 16.1

How Machine Learning System Design covers it

How deep it goes
Defines common fixes and escalation paths, but leaves broader failure learning to retrospectives and postmortems.
Examples
Uses instance scaling during spiky customer demand and escalation to vendors or senior leaders during outages.
What it assumes
Assumes access to operational signals and system documentation, while recognizing that privacy policies may restrict production-data access.
What you can do after
Operate/configure recurring incidents by following repair recipes, scaling guidance, and explicit escalation paths.
How it connects
Connects on-call rotations with logs, metrics, alerts, dashboards, configuration, source documentation, retrospectives, and postmortems.

Also explained in

Compare how 3 books explain runbook →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machine Learning System Design. The section links point to the parts of the book it draws on.