Running stateful workloads
Everything we’ve tried so far with Kubernetes has not answered one important question – what do we do if we need to persist the application state between pod restarts? Data written on a container filesystem is not persisted by default. If you just take a deployment spec from the recent examples with Nginx and replace the image with PostgreSQL, that won’t be enough. Technically, your pod with PostgreSQL will come up, and the database will run, but any data written to that database instance won’t survive a pod restart. But, of course, Kubernetes has something to offer for stateful applications too.
As you hopefully remember from Chapter 4, Exploring Container Runtimes, Interfaces, and Service Meshes, Kubernetes has a Container Storage Interface or CSI that allows you to integrate various storage solutions into a K8s cluster. In order to augment Pods with external storage, we need volumes that can be dynamically provisioned via...
