Raspberry Pi k3s fleet — Rancher installation sequence¶
Parent runbook: How to provision k3s, Longhorn, and Rancher on a Raspberry Pi fleet. Strategy: Rancher — role and timing. External: Install Rancher on Kubernetes (pin chart version at deploy time).
Phase stance¶
| Phase | Rancher |
|---|---|
| P0 | Skip by default—use kubectl, k9s, or similar. |
| P1 | Optional only if k3s, Longhorn, and one stateful app are already stable. |
| Later | Treat Rancher as a critical workload: ingress, TLS, RBAC, upgrades, backups. |
Mandatory — before Helm (whenever you install)¶
- Ingress controller plan exists (bundled Traefik vs NGINX vs other) and you know the hostname Rancher will use.
- TLS strategy chosen: cert-manager + ACME, corporate certs, or lab DNS—not permanent browser click-through for production.
- Resource headroom: Rancher is not free; do not stack it on a maxed Pi also running etcd and Longhorn without measurement.
Mandatory — installation outline¶
Follow Rancher docs for Helm repo add, namespace, ingress options, and bootstrap password.
- Add Helm repo; pick a chart version compatible with your Kubernetes minor.
- Install into
cattle-system(or documented namespace) with explicit hostname, replica count (often one on Pi fleets), and ingress class. - Complete first-login password rotation and document upgrade procedure.
Optional (HA / scale)¶
- Multiple Rancher replicas with anti-affinity needs more RAM than a toy Pi budget—validate in staging.
- Downstream cluster registration for other k3s clusters—later concern for multi-site farms.
Resource warning¶
mermaid
flowchart LR
Pi[Single Pi RAM budget]
Etcd[etcd]
LH[Longhorn]
R[Rancher]
Pi --> Etcd
Pi --> LH
Pi -.->|often too tight| R
If this diagram hurts, defer Rancher or move it to stronger hardware.