k8s

A quick writeup for a longstanding issue I’ve had that’s rendered CephFS-backed PersistentVolumes useless on my home K8S cluster. libceph on worker nodes complains about mons speaking the wrong protocol and that wasn’t a red herring, it was the key to the problem all along.

A whirlwind dive into kube LoadBalancer services, and a lesson in how things are not always what they seem.