Lab: Kubernetes basics with minikube

Aim and what you need

The aim is to learn the core Kubernetes objects by using them: Pods (one or more containers), Deployments (which keep a set number of pods running) and Services (a stable address in front of pods). A local cluster costs nothing and you can break it freely.

• A computer with at least 2 CPUs, 4 GB of free RAM and 20 GB disk.
• Docker installed (see your first Docker container), or a supported hypervisor.
• minikube and kubectl, installed from the official Kubernetes and minikube instructions for your operating system.

Steps

1. Start the cluster with the Docker driver and check the node is Ready.
2. Create a Deployment called hello using the nginx image.
3. Scale it to three replicas and watch the pods appear.
4. Expose it as a NodePort Service and open the URL minikube gives you.
5. Delete one pod and watch a replacement start.
6. Export the Deployment as YAML (YAML Ain't Markup Language) to see the declarative form.

minikube start --driver=docker --cpus=2 --memory=4096
kubectl get nodes
kubectl create deployment hello --image=nginx:alpine
kubectl get deployments,pods -o wide
kubectl scale deployment hello --replicas=3
kubectl get pods -w
kubectl expose deployment hello --type=NodePort --port=80
kubectl get service hello
minikube service hello --url
kubectl delete pod <one-of-the-hello-pod-names>
kubectl get pods
kubectl describe deployment hello
kubectl get deployment hello -o yaml
kubectl logs deployment/hello

How to verify and troubleshoot

Three pods should show Running, and the Service should list three endpoints (kubectl get endpoints hello). The URL from minikube service should show the nginx page. After you delete a pod, a new one with a different name should appear within seconds, because the Deployment's controller sees only two pods where you asked for three.

• Pod stuck in ImagePullBackOff: check the image name and your internet connection; kubectl describe pod shows the exact error.
• minikube will not start: make sure Docker is running and you gave enough memory; minikube logs helps.
• Service URL does not load: on Docker Desktop the minikube service command may need to keep running in its terminal to hold the tunnel open.
• kubectl talks to the wrong cluster: run kubectl config current-context; it should say minikube. This matters if you also have work clusters configured.

Clean-up and next steps

kubectl delete service hello
kubectl delete deployment hello
minikube stop
minikube delete

minikube delete removes the whole cluster and frees the disk space. Next, try writing the Deployment as a YAML file and applying it with kubectl apply -f, then explore Kubernetes RBAC and network policies. If you later try a managed cloud cluster, set a budget first and delete the cluster as soon as you finish, because managed control planes and worker nodes charge by the hour.