Upgrading a Three-Node Talos Cluster One Node at a Time, Run as a Disaster Drill

Overview

A Talos upgrade on a homelab is routine until the day the data on it matters. This guide runs one as if it already did: snapshot etcd, give Longhorn a backup target and prove a restore on a throwaway volume, then upgrade three control-plane nodes one at a time[^1] with the checks between that keep quorum and storage intact. The versions are incidental. The drill is the point, and the backup path it forces into existence is what the cluster's backup series later builds on.