Kubernetes Cluster (kind)

Local Kubernetes for learning + experimentation. No cloud account needed.

Step 1: Install kind

kind is the lightest path to a real Kubernetes cluster on your laptop; it boots Kubernetes in Docker, no cloud account needed.

• macOS. brew install kind.
• Go. go install sigs.k8s.io/kind@latest if you already have a Go toolchain.
• Verify. kind version prints the client version; mismatch with kubectl is rarely a problem at this stage.
• Prereq. Docker Desktop or a Docker-compatible runtime running locally; kind shells out to it.

Step 2: Create cluster

One command brings up a single-node control plane. Boot is roughly 60 seconds depending on whether the node image is cached.

• Create. kind create cluster --name learn spins the cluster and switches your kubectl context to it.
• Wait. First-time pulls the node image; subsequent runs are seconds.
• Verify. kubectl get nodes shows one node in Ready state.
• Multi-node. Pass a config file with worker entries when you outgrow single-node; the rest of this tutorial does not need it.

Step 3: Deploy app

The smallest possible workload is an off-the-shelf nginx image; it proves the cluster can pull, schedule, and run a container.

• Create. kubectl create deployment nginx --image=nginx creates the Deployment and ReplicaSet.
• Inspect. kubectl get pods lists the single pod; kubectl describe pod <name> shows the events if it is stuck.
• Logs. kubectl logs deploy/nginx tails the container output; nginx logs requests once exposed.
• Scale. kubectl scale deploy/nginx --replicas=3 proves the scheduler is doing real work.

Step 4: Expose service

The pod runs but has no ingress until a Service object fronts it; port-forward then lets your laptop reach the cluster IP.

• Service. kubectl expose deployment nginx --port=80 creates a ClusterIP Service in front of the pods.
• Port-forward. kubectl port-forward svc/nginx 8080:80 tunnels local 8080 to the service.
• Verify. Visit http://localhost:8080; the nginx welcome page is your proof of life.
• Cleanup. kind delete cluster --name learn tears the whole thing down when you are done.

Antipatterns

• kind in production. Wrong tool; for learning only.
• Skipping port-forward. No way to see the result.
• Running without cleanup. Resources accumulate.

What to do this week

Three moves. (1) Run the tutorial end-to-end on your own laptop / sandbox. (2) Apply the pattern to one production workload. (3) Document the variations you needed; share with the team.