kind — Run Local Kubernetes Clusters in Docker

What it is

kind (Kubernetes IN Docker) is a tool for running local Kubernetes clusters where each cluster node is a Docker container. It was originally designed for testing Kubernetes itself but has become a standard tool for local Kubernetes development, CI/CD pipelines, and integration testing. kind is a CNCF project maintained by the Kubernetes SIG.

kind is for developers and DevOps engineers who need ephemeral Kubernetes clusters for testing. If you want to test Helm charts, Kubernetes operators, or multi-node cluster behavior without cloud costs, kind provides clusters in seconds.

How it saves time or tokens

kind creates a fully functional Kubernetes cluster in under a minute using only Docker. No VM provisioning, no cloud API calls, no waiting for node boots. Multi-node clusters (control plane plus workers) are defined in a simple YAML config. Clusters are disposable -- create for a test, delete when done. For CI/CD, kind runs inside any environment that has Docker, making Kubernetes testing portable across CI providers.

How to use

1. Install kind: brew install kind (macOS) or go install sigs.k8s.io/kind@latest.
2. Create a cluster: kind create cluster.
3. Use kubectl as normal: kubectl get nodes.

Example

# Install kind
brew install kind

# Create a simple cluster
kind create cluster --name my-test

# Create a multi-node cluster with config
cat <<EOF | kind create cluster --config=-
kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
nodes:
- role: control-plane
- role: worker
- role: worker
EOF

# Load local Docker images into the cluster
kind load docker-image my-app:latest --name my-test

# Use kubectl
kubectl get nodes
kubectl apply -f deployment.yaml

# Delete when done
kind delete cluster --name my-test

Related on TokRepo

• DevOps AI tools -- infrastructure and container management tools
• Docker MCP -- Docker management via Model Context Protocol

Common pitfalls

• Trying to use kind for production workloads. kind is designed for testing and development. It runs on a single machine and lacks the reliability features needed for production Kubernetes.
• Not loading local images into the cluster. kind nodes are Docker containers with their own image cache. Use kind load docker-image to make local images available inside the cluster.
• Running out of Docker resources with large clusters. Each kind node is a Docker container consuming CPU and memory. Limit node count and resource requests for local development.