0 / 15 lessons — 0%
Lesson 02 / 15

Architecture: control plane vs. nodes

Every cluster splits into two kinds of machines: the control plane (the brains) and worker nodes (the muscle, where your actual app containers run).

CONTROL PLANE API server etcd scheduler controller manager everyone talks only to the API server WORKER NODES node 1 kubelet pod pod container runtime node 2 kubelet pod container
Everything flows through the API server. Nothing else talks to etcd or the nodes directly.
ComponentJob
API serverThe front door — every kubectl command, every controller, everyone talks to this
etcdThe cluster's database — the entire desired & actual state lives here
SchedulerDecides which node a new pod should run on, based on resources and constraints
Controller managerRuns the reconciliation loops — "actual replicas = 2, desired = 4, make 2 more"
kubeletRuns on every node — talks to the API server, starts/stops containers as instructed
Container runtimecontainerd or CRI-O — actually pulls images and runs containers, kubelet's helper
Try it yourselfIf you have a cluster handy (even a local one like minikube or kind), run kubectl get nodes and kubectl cluster-info. You're looking at exactly the two halves in the diagram above.