Introduction

Managing failure is the cornerstone of cloud-native. We build autonomous services that limit the blast radius when they do fail and continue to operate when other services fail. We decouple deployment from release, and we control the batch size of each deployment so that we can easily identify the problem when a deployment does go wrong. We shift testing to the left into the continuous deployment pipeline to catch issues before a deployment, as well as all the way to the right into production, where we continuously test the system and alert on anomalies to minimize the meantime to recovery. The recipes in this chapter demonstrate how to design a service to be resilient and forgiving in the face of failure so that transient failures are properly handled, their impact is minimized, and the service can self-heal.