Skip to content

Demo lab

Every study guide has a demo for each competency: commands you paste into a terminal and a description of what you should see. They all run on one local cluster built with kind (Kubernetes in Docker). Set it up once; it takes about five minutes.

A Linux, macOS, or Windows (WSL2) machine with Docker, 4 CPUs, and 8 GB of memory for Docker. Some CKS demos (AppArmor, Falco, WireGuard) need a Linux host kernel; they say so.

Tool Version Used for
kind v0.33.0 The cluster
Kubernetes (kind node image) v1.37.0 The cluster
kubectl v1.37.2 Everything
cloud-provider-kind v0.12.0 LoadBalancer Services, Ingress, and Gateway API (bundles Gateway API v1.6.1)
Helm v4.3.2 Charts (metrics-server, podinfo, Falco, Cilium)
jq, curl, openssl any recent Reading output
cosign v3.1.4 CKS supply chain demos
Trivy v0.76.2 CKS image and SBOM demos
kube-linter v0.8.3 CKS static analysis demo

Versions were current on 17 November 2026. kind v0.33.0 defaults to Kubernetes v1.37.0; the demos pin the node image by digest so the cluster you build matches the one they were written for.

On Linux (amd64). For macOS and arm64, use the matching file names from each project’s release page.

Terminal window
# kind
curl -Lo ./kind https://kind.sigs.k8s.io/dl/v0.33.0/kind-linux-amd64
chmod +x ./kind && sudo mv ./kind /usr/local/bin/kind
# kubectl, checked against its published SHA-256
curl -LO https://dl.k8s.io/release/v1.37.2/bin/linux/amd64/kubectl
curl -LO https://dl.k8s.io/release/v1.37.2/bin/linux/amd64/kubectl.sha256
echo "$(cat kubectl.sha256) kubectl" | sha256sum --check
sudo install -m 0755 kubectl /usr/local/bin/kubectl
# Helm
curl -LO https://get.helm.sh/helm-v4.3.2-linux-amd64.tar.gz
tar -xzf helm-v4.3.2-linux-amd64.tar.gz && sudo install -m 0755 linux-amd64/helm /usr/local/bin/helm
# cloud-provider-kind (needs Go), or download a binary from its v0.12.0 GitHub release
go install sigs.k8s.io/cloud-provider-kind@v0.12.0
sudo install ~/go/bin/cloud-provider-kind /usr/local/bin

One control plane and two workers, so scheduling, draining, and network policy demos have somewhere to move pods.

Terminal window
cat > lab.yaml <<'EOF'
kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
nodes:
- role: control-plane
- role: worker
- role: worker
EOF
kind create cluster --config lab.yaml --image kindest/node:v1.37.0@sha256:a1ed56cfb0e7b93589bdf97c8cd566405a265939e3620fc4f5de89adff580ae5
kubectl get nodes

The nodes are kind-control-plane, kind-worker, and kind-worker2, and the context is kind-kind. Each node is a Docker container, so docker exec kind-worker ... gives you a shell on a node, the way SSH would on a VM.

In a second terminal, start it and leave it running. It gives LoadBalancer Services an IP, and implements Ingress (class cloud-provider-kind) and the Gateway API (GatewayClass cloud-provider-kind), installing the Gateway API CRDs itself.

Terminal window
cloud-provider-kind # on Linux
sudo cloud-provider-kind # on macOS and WSL2; on Windows use an administrator shell

Load balancer IPs sit on Docker’s kind network, which your Mac or Windows host can’t reach directly. The demos call them with docker exec kind-control-plane curl ..., which works everywhere.

kubectl top and the autoscaling demos need it. kind’s kubelets use self-signed serving certificates, so metrics-server must skip verifying them; never do that on a real cluster.

Terminal window
helm repo add metrics-server https://kubernetes-sigs.github.io/metrics-server/
helm install metrics-server metrics-server/metrics-server --version 3.14.0 -n kube-system \
--set 'args={--kubelet-insecure-tls}'
kubectl -n kube-system rollout status deployment/metrics-server
kubectl top nodes
Terminal window
kubectl config current-context # kind-kind
kubectl get nodes # three nodes, Ready
kubectl get gatewayclass,ingressclass # cloud-provider-kind
kubectl top nodes # numbers, not an error
  • One namespace per demo. Most demos start by creating a namespace and switching your context to it, and end by switching back to default and deleting it. Run the clean-up step before the next demo.
  • Images from registry.k8s.io. The demos use the Kubernetes project’s own test images, pinned by tag: registry.k8s.io/e2e-test-images/agnhost:2.66.1 (a small HTTP server with a shell), registry.k8s.io/e2e-test-images/busybox:1.37.0-2, registry.k8s.io/e2e-test-images/nginx:1.28.0-2, and registry.k8s.io/kubectl:v1.37.2. They avoid Docker Hub pull limits. The few exceptions say so.
  • Separate clusters for cluster-level changes. Demos that need special API server flags (audit logging, encryption at rest), a different CNI (Cilium), or three control plane nodes create their own kind cluster and delete it at the end.
  • If something breaks, start again. kind delete cluster and the steps above give you a clean lab in a few minutes.

Every manifest in every demo was validated against the Kubernetes v1.37.2 OpenAPI schema with unknown fields treated as errors, and Gateway API objects against the CRDs bundled with cloud-provider-kind v0.12.0. Every kubectl subcommand and flag was checked against the kubectl v1.37.2 source. Image tags come from the Kubernetes v1.37.2 test image manifest, and third-party versions and commands from each project’s release and documentation.

Terminal window
kind delete cluster