postmortem
A written review after an incident that explains impact, timeline, contributing conditions and the follow-up actions that will change the system.
In more detail
A postmortem is for learning, not blame. It is complete only when its follow-up actions have owners, due dates and success measures and are tracked like other engineering work.
Related terms
- blameless postmortem — A postmortem that asks which conditions made an outcome possible rather than who caused it.
- action item — A specific, owned and verifiable change that a postmortem commits to, such as adding an alert or a test.
- incident timeline — A timestamped record of what was observed, decided and done during an incident, from first symptom to recovery.
- Site Reliability Engineering book — The book in which Google engineers describe how they run production systems, free to read online, with an appendix example postmortem.
- blameless culture — Reviewing failures by examining systems, tools and incentives rather than blaming the individuals involved.
- contributing factors — The conditions that together allowed an incident to happen or made it worse, rather than a single root cause.
- follow-up action — A concrete change agreed in a postmortem, with one owner, a due date and a success measure.
Taught in
- Error Budgets — DevOps & Site Reliability Engineering
- The Incident Lifecycle — DevOps & Site Reliability Engineering
- The Blameless Postmortem — DevOps & Site Reliability Engineering
- Writing A Postmortem: The SRE Appendix — DevOps & Site Reliability Engineering
- Observability Versus Monitoring — Full-Stack Observability: Metrics, Tracing & Logging
- The Blameless Postmortem — Full-Stack Observability: Metrics, Tracing & Logging
- A Worked Example Postmortem — Full-Stack Observability: Metrics, Tracing & Logging