azlocal-expand-2node-switchless

Upgrade a single-node Azure Local cluster to a 2-node switchless cluster — repeatably.

Adding a second node to an Azure Local (Azure Stack HCI) cluster is a fiddly, easy-to-get-wrong operation: you have to identify the right physical host, confirm the storage interconnect, pick a witness, and step around a pile of AKS-Arc / Azure Local gotchas. This repo turns that into a spec-driven, checklist-backed procedure so the next time takes minutes of decisions instead of hours of trial and error.

Companion repo: azlocal-new-2node-switchless (build a new 2-node switchless cluster).

What "switchless" means here

In a 2-node Azure Local cluster you can skip a top-of-rack switch for storage/cluster traffic by directly cabling the two nodes (back-to-back on the storage NIC pair) and using a cloud or file-share witness for quorum. Cheaper, simpler, fewer moving parts — ideal for small edge/lab clusters. This repo assumes (and validates) that switchless layout.

Flow

1. discover/   inventory.sh  -> see your hosts (CPU/RAM/GPU/serial/health)
               match-clusters.sh -> which host belongs to which cluster (don't expand a node into itself!)
2. specs/      copy example.yaml -> describe: existing cluster + the new node + witness
3. validate    (in RUNBOOK) host reachable, storage cross-cabled, hardware homogeneity, witness reachable
4. build/      upgrade-to-2node.sh -> drives the az / Azure Local add-node + extends the AKS-Arc nodepool
5. RUNBOOK.md  the manual physical/portal steps that stay hands-on, as a precise checklist

Quickstart

# 1. inventory your hosts (creds via env — never hardcoded)
IDRAC_USER=root IDRAC_PASS="$YOUR_IDRAC_PASS" ./discover/inventory.sh 192.168.10 2 12

# 2. confirm the target cluster's current node (so you add a DIFFERENT host)
./discover/match-clusters.sh my-cluster:my-rg

# 3. write your spec
cp specs/example.yaml specs/my-cluster.local.yaml   # *.local.yaml is gitignored
$EDITOR specs/my-cluster.local.yaml

# 4. follow RUNBOOK.md; the scriptable parts:
SPEC=specs/my-cluster.local.yaml ./build/upgrade-to-2node.sh

Honest scope

The az / stack-hci / AKS-Arc software steps are scripted. The physical steps (racking, cabling the storage interconnect, the initial Azure Local OS image on bare metal) stay manual — but RUNBOOK.md captures them as an exact checklist, including the gotcha catalog (lib/gotchas.md).

Contributing

PRs welcome — this is community infra tooling (MIT). See CONTRIBUTING.md. Never commit secrets, real iDRAC creds, internal IPs, service tags, or subscription IDs (*.local.yaml and .env are gitignored for that reason).