Deploying a controller+node service across every node in the cluster is a valid approach for supporting local storage via CSI. This should be called out more clearly in the specification as a valid architecture. It's come up in several past conversations and has surprised some folks.
Deploying a controller+node service across every node in the cluster is a valid approach for supporting local storage via CSI. This should be called out more clearly in the specification as a valid architecture. It's come up in several past conversations and has surprised some folks.