Kubernetes

Kubernetes is overkill for most small applications. You likely need Kubernetes if: you're running 50+ microservices that need independent scaling and deployment; you have complex traffic routing requirements; you need fine-grained resource allocation across hundreds of services; or you're building a multi-tenant platform product. For most startups and single-product teams, managed PaaS (Heroku, Render, Railway, Fly.io) or serverless (AWS Lambda, Google Cloud Run) deliver 80% of the operational benefits with 10% of the complexity. The operational overhead of self-managed Kubernetes (or even the learning curve of managed K8s) is significant. 'We use Kubernetes' is a red flag for many early-stage startups where a simple Dockerfile on a $20/month DigitalOcean droplet would suffice.

Docker and Kubernetes solve different problems. Docker runs containers on a single machine — it creates, starts, stops, and manages containers on one server. You use Docker to build an image (Dockerfile), run it locally, and push it to a registry. Kubernetes orchestrates containers across many machines — it decides which of 100 nodes to place your container on, restarts it if it crashes, scales from 2 to 20 replicas under load, and distributes traffic across healthy instances. The typical workflow: Docker to build and test the image, Kubernetes to run it at scale in production. Docker Compose is Docker's limited orchestration for local multi-container development; it doesn't replace Kubernetes for production at scale.

kubectl is the Kubernetes command-line interface — the primary tool for interacting with Kubernetes clusters. Common kubectl commands: `kubectl get pods` (list running containers), `kubectl apply -f deployment.yaml` (deploy or update resources from a YAML file), `kubectl logs pod-name` (stream container logs), `kubectl exec -it pod-name -- bash` (open a shell inside a running container), `kubectl scale deployment/app --replicas=5` (change replica count), `kubectl rollout status deployment/app` (watch a rolling update), and `kubectl port-forward pod-name 8080:80` (forward local port to pod for debugging). kubectl uses kubeconfig files (~/.kube/config) storing cluster credentials and context. Tools like k9s, Lens, and Rancher Desktop provide GUI alternatives to kubectl for teams who prefer visual interfaces.