Internal - Kubernetes Container Lab

Internal admin lab — not for public catalog

This lab is the long-lived test bench for the vm-ide-k8s flavor. It exists to exercise the full vm-ide K8s path end-to-end so we catch regressions in any of the moving parts before they hit user-facing K8s labs:

• The vm-ide-k8s gallery image (kind, kubectl, helm, kindest/node pre-pull, sysctl tuning baked at host level)
• The content-image sidecar pattern (cloud-init extracts src/ → workspace and provision/ → host bootstrap dir)
• The host-side provision/setup.sh execution (creates the kind cluster + writes an in-network kubeconfig before the IDE container starts)
• The lab CLI --k8s scaffold (Dockerfile FROM, COPY provision, starter setup.sh, template branch wiring)

How to use

1. Launch this lab from the admin panel (or run lab ide up locally for the off-platform sim).
2. Expected lab-boot time is comparable to a vm-ide-docker lab (~2 min). Anything materially slower means the kind cluster create on the host is the regression.
3. Inside the IDE terminal, sanity check:
   • kubectl get nodes — should show the kind control-plane node Ready.
   • kubectl -n kube-system get pods — system pods all Running.
   • k9s — TUI opens and lists cluster resources.

What to check when something breaks

• Lab fails to launch: check the lab session's cloud-init log (worker pulls it on teardown). Failure in extract /root/lab/provision/ → content-image scaffold drift. Failure in kind create cluster → host bake regression (sysctl, kindest/node pre-pull, or kind binary).
• Cluster up but kubectl from inside IDE can't reach it: the in-network kubeconfig wiring (/srv/lab-kube/config mount → /home/labs/.kube/config) is broken. Check the IDE container's docker run args.
• IDE container missing kubectl/helm/k9s: the cloudlearn-labs:ide-k8s image is out of date. Rebuild + re-bake vm-ide-k8s gallery image.