Deployments

Graceful Shutdown and Connection Draining for Cloud Services

Cloud services are stopped routinely: deployments replace instances, autoscalers remove capacity, hosts reboot, and schedulers reschedule workloads. If a process exits immediately when it receives a termination signal, in-flight requests can fail even when the service is otherwise healthy. Graceful shutdown is a small lifecycle protocol: stop taking new work, allow useful work already in progress to finish within a deadline, then release resources and exit. Separate readiness from process health A process that is shutting down may still be alive but should no longer receive new traffic.