Lesson 3: The anatomy of a cluster
A cluster is a set of machines (servers) that Kubernetes manages as if they were one big computer. But not every machine does the same job, and that is the core idea of this lesson: there is a difference between "who decides" and "who runs". The deciding part is called the control plane: it receives
A cluster is like a restaurant: at the front a manager takes orders and decides who cooks (the control plane), and in the kitchen cooks actually make the food (the nodes). You only talk to the front.
- Cluster
- The set of machines (servers) Kubernetes manages as a single unit. Some decide (the control plane) and some actually run the applications (worker nodes).
- Control Plane
- The deciding part. The set of management components that receive requests, store the desired state, schedule work onto machines, and run reconciliation to close gaps.
- Worker Node
- The running part. A machine in the cluster on which Pods actually run. Each node runs an agent called kubelet and a container runtime.
- API Server
- The cluster's "front door". The only component kubectl talks to: it validates requests, stores them in etcd, and triggers the other control-plane components.
- kubectl
- The command-line tool you use to talk to the cluster. Every command it sends goes to the API Server, never directly to the nodes or Pods.
