Graceful Shutdown for a Node Server

Make a Node server shut down cleanly when the platform sends SIGTERM. In-flight requests finish or time out. The process does not hang forever. SERVER [FRAMEWORK] What is in flight: [HTTP, A POOL, A QUEUE WORKER] Platform stop grace: [SECONDS] Database pool: [LIBRARY] DELIVER 1) On SIGTERM: stop accepting new connections, mark readiness as failing so the load balancer leaves, wait for in-flight requests up to a timeout shorter than the platform grace, then close the pool and exit 0. 2) On the timeout: exit non-zero so we can see it, after attempting to close. 3) The worker variant: stop taking new jobs, finish the current one or return it to the queue if the platform is about to kill us. Do not ack a job we did not finish. 4) A local way to test: send the signal and show which log line we expect. 5) The bug to avoid: a server that listens for SIGTERM and then never exits because a handle is open. Name the handle (pool, timer, stream). Do not add a second listener that calls process.exit immediately and skips the drain.