High Availability and Services with Kubernetes

In our previous blog, Getting Started with a Local Deployment, we deployed an Nginx pod to a standalone (single-node) Kubernetes cluster. This pod was bound to a specified node. If the pod were to fail unexpectedly, Kubernetes (specifically, the Kubelet service) would restart the pod. By default, pods have an ‘Always’ restart policy, but only to the node that it is first bound; it will not be rebound to another node. This means of course that if the node fails then pods will not be rescheduled elsewhere.