My 3-node Talos homelab had no persistent storage, so anything with data, like Grafana or a database, had nowhere to live.
Longhorn keeps every volume on two of the three nodes, so one node can go down without losing data, and it rebuilds the missing copy on the third node.
It also takes hourly snapshots, and backups to S3 are optional. The Longhorn UI runs on my LAN and over Tailscale.