Enabling high availability in a Kubernetes cluster