阅读中文版本

Test Graceful Shutdown Signal Handling

Assert that a service stops accepting new work, drains in-flight requests, and exits within the shutdown deadline on SIGTERM.

const child = spawn('node', ['server.js']);
const slow = fetch('http://localhost:3000/slow');
await new Promise(r => setTimeout(r, 200));

child.kill('SIGTERM');
const res = await slow;
expect(res.status).toBe(200);
await once(child, 'exit');

What this recipe does

Write an integration test that sends SIGTERM and verifies draining behaviour.

Deployments cut off in-flight requests because the process exits immediately on SIGTERM instead of draining.

Prerequisites

• A service that handles SIGTERM
• A test harness that can send signals
• A slow endpoint to observe draining

Procedure

1. Start the service and issue a slow request.

const child = spawn('node', ['server.js']);
const slow = fetch('http://localhost:3000/slow');
await new Promise(r => setTimeout(r, 200));

Expected result: A request is in flight when the shutdown begins.

2. Send SIGTERM and assert the request completes.

child.kill('SIGTERM');
const res = await slow;
expect(res.status).toBe(200);
await once(child, 'exit');

Expected result: The in-flight request finishes and the process exits, proving graceful draining.

How to confirm it worked

Run the commands above and confirm every expected result in order. If a step does not produce the stated output, use the troubleshooting notes before continuing. A passing run shows the in-flight request completes with 200 and the process exits within the configured grace period.

Troubleshooting

Process exits before draining — Register the signal handler early and stop the listener before draining.

Shutdown hangs forever — Add a hard deadline timer that forces exit after the grace period.

Limitations

• Does not cover load-balancer connection draining.
• Signal semantics differ between platforms.
• Requires the slow endpoint to expose the behaviour.