Hands-On Microservices with Kubernetes

Container failures

Kubernetes runs containers inside pods. If a container dies for whatever reason, Kubernetes will detect it and restart it right away by default. The behavior of Kubernetes can be controlled by the restartPolicy file of the pod spec. The possible values are Always (default), OnFailure, and Never. Note that the restart policy applies to all the containers in the pod. There is no way to specify a restart policy per container. This seems a little short-sighted as you may have multiple containers in a pod that require a different restart policy.

If a container keeps failing, it will enter a CrashOff. Let's see this in action by introducing an intentional error to our API gateway:

import os from api_gateway_service.api import ...