First cluster comes up with ZERO workers — clusterctl overrides WORKER_MACHINE_COUNT to 0; the effective default must be 1

Symptom

Follow "Deploy your first cluster" verbatim and you get a control-plane-only cluster — no worker node, silently. clusterctl describe shows no MachineDeployment branch, and:

machinedeployment/<cluster>-md-0   DESIRED=0  CURRENT=0  READY=0
machineset/<cluster>-md-0-...      DESIRED=0

The user reasonably expects a worker to appear (the base template ships a MachineDeployment). It never does, because it was created with replicas: 0.

Root cause

clusterctl generate cluster has a --worker-machine-count flag that defaults to 0 and is always injected as the WORKER_MACHINE_COUNT variable:

--control-plane-machine-count int   (default 1)
--worker-machine-count int          (default 0)

Our base template declares replicas: ${WORKER_MACHINE_COUNT:=1} — but that :=1 default is dead code, because clusterctl stamps WORKER_MACHINE_COUNT=0 over it whenever the flag is omitted. So unless the user explicitly passes --worker-machine-count N, they get zero workers. (--control-plane-machine-count defaults to 1, which is why the control plane looks fine and hides the asymmetry.)

Verified live this session: base-template cluster came up with 1 control-plane Machine and an md-0 MachineDeployment at replicas: 0; kubectl scale machinedeployment <cluster>-md-0 --replicas=1 immediately created the worker Machine.

Desired behaviour

A user following the getting-started guide verbatim should get 1 control plane + 1 worker, both reaching Ready. The effective default must be 1 worker, not 0.

Fix

Since clusterctl owns the value (the template default can't win), the fix is in the docs/examples, not the template :=1:

  1. Every clusterctl generate cluster example in getting-started + guides must pass --worker-machine-count 1 explicitly (or a higher number where the guide wants it). This is the real "default 1". 
    clusterctl generate cluster my-cluster --kubernetes-version <ver> \
  --infrastructure vco --worker-machine-count 1 | kubectl apply -f -
  2. docs/reference/templates.md: document that WORKER_MACHINE_COUNT is effectively required — clusterctl overrides the template's :=1 to 0 when the flag is omitted — so it belongs in the "must supply" story, not the "defaults to 1" one. Fix the current wording that claims a default of 1.
  3. first-cluster.md verification: show the expected kubectl get nodes output with both a control-plane and a worker node, so a missing worker is caught immediately instead of read as normal.
  4. Optionally note the day-2 lever: kubectl scale machinedeployment <cluster>-md-0 --replicas=N (and that --worker-machine-count seeds the initial count).

Acceptance

Running the first-cluster guide exactly as written produces a cluster with 1 control-plane and 1 worker node, both Ready — no manual scale step required.

Related: #14 (closed) (CNI never lands) and #13 (closed) (hardcoded image ids) — same "Deploy your first cluster must be smooth, nothing can go wrong" theme.