Design a control plane for cloud clusters with host health checks

Design the management layer of a cloud platform that runs clusters of hosts (VMs or physical machines). It takes requests to create, resize and change clusters, decides which hosts belong to which cluster, and keeps driving every cluster toward the configuration that was asked for. It also watches host health and replaces hosts that fail.

Cover the API, where desired and actual state live, the reconcile loop, how health is detected (heartbeats, probes) and acted on, what happens when the control plane itself fails, and how it scales to many clusters.