Convergence

Convergence compares the configuration Control expects with the current state it can observe and repairs supported drift.

States

Common states include CONVERGED, OUT_OF_STATE, WAITING, REPAIRING, VERIFYING, NEEDS_ATTENTION and UNKNOWN.

Ordinary safe, repairable drift can be repaired without marking an otherwise healthy serving Site as degraded. Attention is reserved for conditions that are unsafe, ambiguous or repeatedly unsuccessful.

Independent repair areas

Filesystem, TLS, runtime, managed application, jobs/workers, Mail and data/runtime dependencies can be evaluated separately so one repair does not hide unrelated healthy state.

Waiting and dependencies

A resource can be WAITING when it depends on another condition, such as application setup or a Python runtime repair. Waiting is not the same as repeatedly failing the dependent job.

Retry and cancellation

When a repair requires operator action, correct the cause and use the available retry/reconcile action. Cancellation stops supported in-progress repair work but does not pretend application side effects can always be undone.