ClusterIP and Service Discovery

Every Pod gets an IP address. That IP address is ephemeral. When the Pod restarts, the IP changes. When a rollout replaces one ReplicaSet with another, the whole set of IPs changes. Building cross-service communication on top of Pod IPs means writing service registries, maintaining retry logic, and pushing health-check concerns into application code that shouldn't have to care. Services exist to take that burden away.

ClusterIP is the default Service type and the foundation that NodePort and LoadBalancer extend. When you run kubectl expose deployment, you get a ClusterIP service. Before you can reason about cross-namespace communication, StatefulSet DNS discovery, or why your frontend can't reach the backend after a rollout, you need a clear mental model of what ClusterIP actually is — and it's not a network interface, a proxy process, or a load balancer in the traditional sense.