← Back to docsRecipe

Runbook generator from past incidents

Turn your incident history into structured, repeatable runbooks that your team can execute under pressure.

Ingredients

  • Postmortem documents from the last 6 months
  • Incident timeline exports (PagerDuty, Opsgenie, or CSV)
  • Monitoring alerts correlated to each incident
  • Access to runbook template (Confluence, Notion, or Markdown)

Method

  1. 1Cluster incidents by root cause category. Group similar failures together so each runbook covers a distinct failure mode.
  2. 2Extract the diagnostic path. For each cluster, list the exact queries, dashboards, and log filters responders used to confirm the issue.
  3. 3Document the mitigation steps in order. Include exact commands, feature flags, and rollback procedures.
  4. 4Add escalation triggers. Define clear thresholds for when to page secondary on-call or escalate to engineering leads.
  5. 5Validate with a tabletop exercise. Walk the runbook against a simulated incident and refine gaps.

Output

A set of Markdown runbooks organized by failure mode, each containing diagnostic queries, step-by-step mitigations, escalation paths, and links to relevant dashboards.

Pro tip: Store runbooks alongside your alert definitions so they surface automatically when an alert fires. Meridian can link alerts directly to runbook steps.