Handle SIGTERM or every deploy is a small outage

If your Node service does not handle SIGTERM, every deploy is a small outage. The fix is a graceful shutdown handler of about fifteen lines.

What happens without it

The pod gets the signal, the process dies immediately, and whatever requests were in flight just end. Users see a 502 and blame the feature you shipped.

The fix

Catch SIGTERM, stop accepting new connections, let in-flight requests finish, close the database pool, then exit. Give it a timeout so a stuck request cannot block the shutdown forever.

const server = app.listen(port);
 
process.on("SIGTERM", () => {
  // Hard stop if something hangs
  const timer = setTimeout(() => process.exit(1), 10_000);
  timer.unref();
 
  // Stop accepting new connections, wait for in-flight requests
  server.close(async () => {
    await db.end(); // close the database pool
    process.exit(0);
  });
});

Keep the timeout shorter than your orchestrator's grace period, so your code decides how the process ends rather than a SIGKILL.

Nobody notices

Nobody notices when you get this right. That is sort of the point.

Takeaways

• Without a SIGTERM handler, every deploy cuts off in-flight requests.
• Stop accepting connections, drain requests, close pools, then exit.
• Add a timeout so a stuck request cannot block shutdown forever.
• It is a small amount of code for a quieter deploy.