mitigation
An action that stops or limits harm to users, such as a rollback or routing traffic around a failure, even before the cause is fixed.
In more detail
Prefer a safe, reversible mitigation, such as a rollback, a kill switch or moving traffic, over a speculative fix. Record mitigations separately from suspected causes, because restoring service often comes before proving why it failed. In an incident timeline, time to mitigate is its own clock, separate from detection and from the final fix.
Related terms
- rollback — Returning a system to a previous known-good version after a bad change.
- kill switch — A control, usually a feature flag, that turns a feature off immediately without a new deployment.
- incident lifecycle — The stages of an incident: detection, declaration, mitigation, end of user impact and closure.
- incident — An unplanned event that degrades or interrupts a service for its users, or harms the security of systems or data.
- incident timeline — A timestamped record of what was observed, decided and done during an incident, from first symptom to recovery.
- postmortem — A written review after an incident that explains impact, timeline, contributing conditions and the follow-up actions that will change the system.
Taught in
- Continuous Delivery Versus Continuous Deployment — DevOps & Site Reliability Engineering
- Decoupling Release From Deployment: Feature Flags — DevOps & Site Reliability Engineering
- On-Call And Incident Response — DevOps & Site Reliability Engineering
- The Incident Lifecycle — DevOps & Site Reliability Engineering
- The Blameless Postmortem — DevOps & Site Reliability Engineering
- On Call And Incident Command — 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