reconciliation loop
also: control loop, operator pattern
A control mechanism in Kubernetes and similar systems that continuously compares the desired state with the actual state and makes corrections to keep them in sync.
A reconciliation loop is a fundamental pattern in declarative systems where a controller repeatedly checks whether the current state of a resource matches its desired state. If they differ, the controller takes action to bring the actual state into agreement with the desired state.
In Kubernetes, for example, if you declare that a Deployment should run 3 replicas of a pod, the Deployment controller's reconciliation loop continuously verifies this. If a pod crashes and only 2 remain, the loop detects this mismatch and automatically creates a new pod to reach the desired count of 3.
This pattern is powerful because it makes the system self-healing: you define the end goal, and the reconciliation loop works continuously to maintain it without requiring manual intervention. The loop typically runs every few seconds and processes events (like pod failures) as they occur.