Self-heal from a Worker Node Failure

When a worker node fails, all Pods running on the node are lost. However, if a Deployment controller manages the Pods, Kubernetes will start replacements on surviving nodes.

If your cluster is on a cloud that implements node pools, the failed node will also be replaced. However, this is a feature of your cloud’s hosted Kubernetes service and not a feature of Deployments—Deployments do not heal nodes; they only heal Pods.

Playground

The steps in this section will use a Multi-node cluster. Run the following widget to create a cluster and deploy five Pods using the deploy.yml file.