The Brain and the Body
A Kubernetes cluster is divided into two distinct roles: the machines that run your workloads, and the machines that manage everything else.
- The Worker Nodes (Body): Virtual machines or physical servers that run your application containers. They do the actual compute work.
- The Control Plane (Brain): A set of specialized components that make decisions about the cluster — what should run, where it should run, and what to do when something goes wrong. In older Kubernetes documentation, these were called "master" nodes, but the term "control plane" is now preferred.
Think of the control plane like the air traffic control tower at a busy airport. The planes (worker nodes) do the actual flying, but without the tower coordinating takeoffs, landings, and routing, the whole system collapses into chaos. The tower doesn't carry passengers — it just makes sure everything runs safely and efficiently.
The control plane makes scheduling decisions, detects and responds to cluster events (like restarting a crashed Pod), and stores all cluster state. If the control plane goes down, your already-running applications may continue operating for a while — but you lose the ability to deploy new workloads, scale existing ones, or recover from failures. No new planes can take off or land.
The four core control plane components are:
- kube-apiserver — the single entry point for all cluster operations
- etcd — the persistent store for all cluster state
- kube-controller-manager — the engine that drives the cluster toward desired state
- kube-scheduler — the component that assigns Pods to Nodes
Each of these runs as a process (typically a Pod itself, on a dedicated control plane node) and communicates exclusively through the API Server.