Implementing a Highly-Available Control Plane

Running a single control plane node is a maintenance window masquerading as a cluster. The moment you drain it for a patch or upgrade, every control loop freezes — no new Pods schedule, no Deployments reconcile, no kubectl commands succeed. Workloads already running keep going, but the cluster is effectively blind.

The CKA exam tests whether you know how to avoid this. The "Cluster Architecture, Installation and Configuration" domain (25% of the exam) covers HA topology choices, the kubeadm bootstrap sequence, certificate distribution, and etcd quorum requirements. More importantly, these are decisions you'll make in production and can't undo.