Kubernetes Primer

Kubernetes is an orchestration platform for running containers. Its defining trait is declarative intent: you declare what you want (desired state), and the system continuously tries to make reality match it.

That continuous convergence is why Kubernetes can self-heal: if a Pod dies, controllers recreate it; if a node disappears, workloads can be rescheduled; if replicas drift, the control plane works to restore them.

A Kubernetes cluster is the whole system: the control plane (API server, etcd, controllers, scheduler) and the worker nodes where Pods run.

kubectl config get-contexts
kubectl config current-context

A node is a VM or physical machine registered into the cluster. Nodes run a container runtime and the kubelet, which talks to the control plane and ensures containers are running as instructed.

kubectl get nodes -o wide

A Pod is one or more tightly coupled containers that share a network namespace and can share storage. Pods are ephemeral: you don’t typically manage individual pods directly in production.

kubectl get pods -A
kubectl get pods -n default

A Deployment manages ReplicaSets, which manage Pods. Deployments provide rollouts, rollbacks, and a stable intent: “run N replicas of this template.”

A Service gives you a stable IP/DNS name and routes to a changing set of backends (Pods). It relies on labels/selectors and endpoint data.

An Ingress defines HTTP routing rules, but it requires an Ingress Controller(like NGINX Ingress or Traefik) to actually enforce them. Ingress is not “built-in routing magic”—it’s a contract interpreted by a controller.