Local Clusters: minikube & kind
minikube and kind let you run a full Kubernetes cluster on your own laptop for learning and testing.
You do not need a cloud account to learn Kubernetes. Two popular tools create a real cluster on your machine:
- minikube — runs a single-node cluster inside a VM or container. Great for beginners.
- kind — "Kubernetes IN Docker"; runs nodes as Docker containers. Fast and popular for CI.
Getting started with minikube
After installing it, one command boots a cluster and points kubectl at it automatically:
minikube startFrom there, every kubectl command talks to your local cluster. When you are done, minikube stop pauses it and minikube delete removes it entirely.
Contexts
A context tells kubectl which cluster and user to use. kubectl config get-contexts lists them and kubectl config use-context NAME switches between, say, your local cluster and a production one.
Example
# Start a local single-node cluster
minikube start
# Confirm it is running
kubectl get nodes
# Open the built-in dashboard in your browser
minikube dashboard
# Stop it when finished
minikube stopWhen to use it
- A developer runs minikube to test a new Helm chart on their laptop before submitting it to the staging cluster.
- A team uses kind in CI to run Kubernetes integration tests against a real API server without cloud infrastructure costs.
- A student uses minikube to learn Kubernetes hands-on locally without requiring a cloud provider account.
More examples
Start and verify minikube cluster
Starts a single-node minikube cluster using the Docker driver and verifies the node is Ready to schedule pods.
minikube start --driver=docker --cpus=2 --memory=4096
kubectl get nodes
minikube statusExpose service with minikube tunnel
Creates a deployment, exposes it as LoadBalancer, then tunnels external traffic through so localhost can reach the service.
kubectl create deployment hello --image=nginx
kubectl expose deployment hello --port=80 --type=LoadBalancer
minikube tunnelMulti-node kind cluster config
Defines a kind cluster with one control-plane and two worker nodes, enabling local testing of multi-node scheduling.
# kind-config.yaml
kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
nodes:
- role: control-plane
- role: worker
- role: worker
Discussion