k3s cluster bootstrap standard — smart farm / homelab¶
Purpose¶
Canonical expectations for ordering and gates when bringing up a k3s cluster that may host farmOS, Longhorn, and edge workloads—not a duplicate of the full runbook.
Runbook (procedure): How to provision k3s, Longhorn, and Rancher on a Raspberry Pi fleet. Node checklist: Raspberry Pi fleet provisioning standard — smart farm / homelab.
Standard (must hold)¶
| # | Requirement | Rationale |
|---|---|---|
| 1 | One documented bootstrap sequence per environment (lab vs prod)—no ad-hoc re-installs without updating docs | Reproducible recovery |
| 2 | Server before agents; kubeconfig and join token handled as secrets (Secrets and certificates — edge cluster standard) |
k3s join contract |
| 3 | Longhorn only after nodes meet iSCSI + disk prerequisites per Pi standard | CSI prerequisites |
| 4 | StorageClass and default policy agreed before stateful workloads (Kubernetes edge — scheduling, storage…) |
Avoid PVC churn |
| 5 | Backup tracks (logical DB + etcd path) named before declaring “production” (Backup validation cadence) |
Empty clusters are not success |
Phase alignment: Phase 0/1 stays single-server k3s unless Platform decision memo justifies HA.