Control Plane: the brain of the cluster
The Control Plane is the set of components that make decisions: it takes your requests (via kubectl), decides where pods run, and keeps actual state matching desired state.
- kube-apiserver — the only gateway into the cluster
- etcd — the key-value store holding all cluster state
- kube-scheduler — picks a node for unscheduled pods
- kube-controller-manager — controllers that enforce desired state
Node: where workloads actually run
Each node runs kubelet (talks to the apiserver), kube-proxy (Service networking) and a container runtime (usually containerd).
bash
kubectl apply -f deployment.yaml # 1. apiserver stores the desired state in etcd # 2. scheduler picks a node # 3. kubelet pulls the image and starts the container
Why the boundary matters
A Pending pod usually means a scheduler or kubelet problem. A kubectl connection failure points at the Control Plane. Knowing which side is broken makes debugging far faster.
Rule of thumb: the Control Plane decides what runs where; Nodes just run it correctly.