management¶

Purpose¶

Stable names for cluster control on the farm edge stack: k3s server (API + embedded etcd in typical single-server patterns), agents, Longhorn control/data plane components as storage roles, and Rancher (if deployed) as multi-cluster management—distinct from application workloads (farmOS, brokers).

Scope¶

In scope: Role boundaries and backup implications at a high level (etcd vs app data).
Out of scope: Cluster endpoint URLs, certificate SANs, or node hostnames.

Knowns¶

k3s and Longhorn roles are documented in dedicated analyses; DR package maps three tracks (app / Longhorn / etcd).
Rancher’s timing and scope are explicitly debated (see Rancher role analysis).

Assumptions¶

etcd recovery without logical app backups does not restore farm records—see restore tiers.

Outstanding unknowns¶

Multi-server k3s topology vs single-server pilot default for your fleet (decision memo captures tradeoffs—not a single answer here).