Graceful Shutdown

Q: How long should preStop sleep be?

3-5 seconds often enough for endpoint propagation before SIGTERM. Tune with load balancer docs.

Q: Does Cloud Run send SIGTERM?

Yes on instance scale-in and revision deploy. Same handler pattern applies.

Q: Lambda needs graceful shutdown?

Short-lived invocations; callbackWaitsForEmptyEventLoop matters more than SIGTERM. See Lambda Handler Patterns.

Q: WebSocket connections?

Close WebSocket server in shutdown sequence; notify clients with close frame. May need longer grace period.

Q: NestJS?

app.enableShutdownHooks() listens for SIGTERM and closes modules implementing OnModuleDestroy.

Q: What exit code on forced shutdown?

Exit 1 on timeout so supervisor logs abnormal termination and alerts fire.

Handle SIGTERM and SIGINT by stopping new work, draining HTTP connections, and closing database pools before exit on Node.js 24.

Recipe

Quick-reference recipe card - copy-paste ready.

import express from "express";
import { Pool } from "pg";
 
const app = express();
const pool = new Pool({ connectionString: process.env.DATABASE_URL });
const server = app.listen(Number(process.env.PORT ?? 3000), "0.0.0.0");
 
let shuttingDown = false;
 
app.get("/ready", (_req, res) => {
  if (shuttingDown) return res.status(503).json({ status: "draining" });
  res.json({ status: "ready" });
});
 
function shutdown(signal: string) {
  shuttingDown = true;
  console.log(JSON.stringify({ event: "shutdown_start", signal }));
 
  server.close(async () => {
    await pool.end();
    console.log(JSON.stringify({ event: "shutdown_complete" }));
    process.exit(0);
  });
 
  setTimeout(() => {
    console.error(JSON.stringify({ event: "shutdown_forced" }));
    process.exit(1);
  }, 30_000).unref();
}
 
process.on("SIGTERM", () => shutdown("SIGTERM"));
process.on("SIGINT", () => shutdown("SIGINT"));

When to reach for this: Every production HTTP service on Kubernetes, ECS, Cloud Run, PM2, or systemd.

Working Example

// src/shutdown.ts
import type { Server } from "node:http";
import type { Pool } from "pg";
 
export type Closable = { close: () => Promise<void> };
 
export function registerGracefulShutdown(
  server: Server,
  resources: Closable[],
  options: { timeoutMs?: number } = {}
) {
  const timeoutMs = options.timeoutMs ?? 30_000;
  let draining = false;
 
  const shutdown = async (signal: string) => {
    if (draining) return;
    draining = true;
    console.log(JSON.stringify({ event: "shutdown_start", signal }));
 
    await new Promise<void>((resolve, reject) => {
      server.close((err) => (err ? reject(err) : resolve()));
    });
 
    for (const r of resources) {
      await r.close();
    }
 
    console.log(JSON.stringify({ event: "shutdown_complete" }));
    process.exit(0);
  };
 
  const forceTimer = setTimeout(() => {
    console.error(JSON.stringify({ event: "shutdown_timeout" }));
    process.exit(1);
  }, timeoutMs);
  forceTimer.unref();
 
  process.on("SIGTERM", () => void shutdown("SIGTERM"));
  process.on("SIGINT", () => void shutdown("SIGINT"));
 
  return {
    isDraining: () => draining,
  };
}
 
// src/main.ts
import express from "express";
import { Pool } from "pg";
import { registerGracefulShutdown } from "./shutdown.js";
 
const pool = new Pool({ connectionString: process.env.DATABASE_URL });
const app = express();
const shutdownCtl = registerGracefulShutdown(
  app.listen(3000, "0.0.0.0"),
  [{ close: () => pool.end() }]
);
 
app.get("/ready", (_req, res) => {
  if (shutdownCtl.isDraining()) {
    res.status(503).json({ status: "draining" });
    return;
  }
  res.json({ status: "ready" });
});

What this demonstrates:

server.close() stops accepting; finishes in-flight requests
pool.end() drains DB connections cleanly
Readiness fails during drain so load balancers stop sending traffic
Forced exit after 30s matches typical terminationGracePeriodSeconds

Deep Dive

Signal Timeline on Kubernetes

1. Pod marked Terminating
2. Endpoints removed from Service (readiness fails)
3. preStop hook runs (optional sleep)
4. SIGTERM sent to container
5. grace period elapses -> SIGKILL

Align app shutdown timeout with pod terminationGracePeriodSeconds (app timeout < grace period).

Fastify Shutdown

import Fastify from "fastify";
 
const app = Fastify();
app.addHook("onClose", async () => {
  await pool.end();
});
 
const close = async () => {
  await app.close();
  process.exit(0);
};
process.on("SIGTERM", () => void close());

app.close() triggers onClose hooks in reverse registration order.

In-Flight Work Beyond HTTP

Resource	Close API
PostgreSQL `pg.Pool`	`pool.end()`
Redis `ioredis`	`redis.quit()`
BullMQ worker	`worker.close()`
Cron `setInterval`	`clearInterval` + await last job

Track background jobs; server.close() alone does not wait for queue workers.

`@godaddy/terminus`

import { createTerminus } from "@godaddy/terminus";
 
createTerminus(server, {
  signals: ["SIGTERM", "SIGINT"],
  healthChecks: { "/health": () => Promise.resolve() },
  onSignal: async () => { await pool.end(); },
});

Bundles health checks and shutdown for Express apps.

Gotchas

No readiness fail on shutdown - LB sends traffic to dying pod. Fix: shuttingDown flag on /ready.
server.close() without timeout - hang forever on keep-alive. Fix: force exit timer; tune keepAliveTimeout on server.
Closing pool before HTTP drain - in-flight requests error. Fix: server.close() first, then pools.
SIGTERM ignored in dev - Ctrl+C sends SIGINT; handle both.
Workers keep consuming after HTTP stops - duplicate processing during deploy. Fix: pause workers on SIGTERM.
PM2 kill_timeout too short - hard kill mid-request. Fix: 30s kill_timeout - PM2 & systemd.

Alternatives

Alternative	Use When	Don't Use When
Manual SIGTERM handler	Full control, any framework	Want batteries-included health + drain
@godaddy/terminus	Express HTTP services	Fastify (use `app.close()`)
Kubernetes only (no app handler)	Never	Production APIs (always handle SIGTERM)
Immediate `process.exit(0)`	Batch jobs	HTTP services (causes 502s)

FAQs

How long should preStop sleep be?

3-5 seconds often enough for endpoint propagation before SIGTERM. Tune with load balancer docs.

Does Cloud Run send SIGTERM?

Yes on instance scale-in and revision deploy. Same handler pattern applies.

Lambda needs graceful shutdown?

Short-lived invocations; callbackWaitsForEmptyEventLoop matters more than SIGTERM. See Lambda Handler Patterns.

WebSocket connections?

Close WebSocket server in shutdown sequence; notify clients with close frame. May need longer grace period.

NestJS?

app.enableShutdownHooks() listens for SIGTERM and closes modules implementing OnModuleDestroy.

What exit code on forced shutdown?

Exit 1 on timeout so supervisor logs abnormal termination and alerts fire.

Zero-Downtime Deploys - rollout + drain
Health & Readiness Probes - readiness during drain
Kubernetes Deployment - preStop and grace period
Runtime Ops Basics - supervision model
Runtime Ops Best Practices - section checklist

Node.js API Documentation - Complete reference for core modules and APIs.
Node.js Best Practices (GitHub) - Curated production best practices.
DigitalOcean Node.js Tutorials - Production-oriented how-tos.
Node.js Design Patterns, 4th Edition - Production-grade application patterns.
Distributed Systems with Node.js - Enterprise-ready distributed systems.

Stack versions: This page was written for Node.js 24.18.0 (Active LTS), npm 10+, TypeScript 5.6+, Express 5, Fastify 5, and NestJS 11.

Recipe

Quick-reference recipe card - copy-paste ready.

import express from "express";
import { Pool } from "pg";
 
const app = express();
const pool = new Pool({ connectionString: process.env.DATABASE_URL });
const server = app.listen(Number(process.env.PORT ?? 3000), "0.0.0.0");
 
let shuttingDown = false;
 
app.get("/ready", (_req, res) => {
  if (shuttingDown) return res.status(503).json({ status: "draining" });
  res.json({ status: "ready" });
});
 
function shutdown(signal: string) {
  shuttingDown = true;
  console.log(JSON.stringify({ event: "shutdown_start", signal }));
 
  server.close(async () => {
    await pool.end();
    console.log(JSON.stringify({ event: "shutdown_complete" }));
    process.exit(0);
  });
 
  setTimeout(() => {
    console.error(JSON.stringify({ event: "shutdown_forced" }));
    process.exit(1);
  }, 30_000).unref();
}
 
process.on("SIGTERM", () => shutdown("SIGTERM"));
process.on("SIGINT", () => shutdown("SIGINT"));

When to reach for this: Every production HTTP service on Kubernetes, ECS, Cloud Run, PM2, or systemd.

Working Example

// src/shutdown.ts
import type { Server } from "node:http";
import type { Pool } from "pg";
 
export type Closable = { close: () => Promise<void> };
 
export function registerGracefulShutdown(
  server: Server,
  resources: Closable[],
  options: { timeoutMs?: number } = {}
) {
  const timeoutMs = options.timeoutMs ?? 30_000;
  let draining = false;
 
  const shutdown = async (signal: string) => {
    if (draining) return;
    draining = true;
    console.log(JSON.stringify({ event: "shutdown_start", signal }));
 
    await new Promise<void>((resolve, reject) => {
      server.close((err) => (err ? reject(err) : resolve()));
    });
 
    for (const r of resources) {
      await r.close();
    }
 
    console.log(JSON.stringify({ event: "shutdown_complete" }));
    process.exit(0);
  };
 
  const forceTimer = setTimeout(() => {
    console.error(JSON.stringify({ event: "shutdown_timeout" }));
    process.exit(1);
  }, timeoutMs);
  forceTimer.unref();
 
  process.on("SIGTERM", () => void shutdown("SIGTERM"));
  process.on("SIGINT", () => void shutdown("SIGINT"));
 
  return {
    isDraining: () => draining,
  };
}
 
// src/main.ts
import express from "express";
import { Pool } from "pg";
import { registerGracefulShutdown } from "./shutdown.js";
 
const pool = new Pool({ connectionString: process.env.DATABASE_URL });
const app = express();
const shutdownCtl = registerGracefulShutdown(
  app.listen(3000, "0.0.0.0"),
  [{ close: () => pool.end() }]
);
 
app.get("/ready", (_req, res) => {
  if (shutdownCtl.isDraining()) {
    res.status(503).json({ status: "draining" });
    return;
  }
  res.json({ status: "ready" });
});

What this demonstrates:

server.close() stops accepting; finishes in-flight requests
pool.end() drains DB connections cleanly
Readiness fails during drain so load balancers stop sending traffic
Forced exit after 30s matches typical terminationGracePeriodSeconds

Deep Dive

Signal Timeline on Kubernetes

1. Pod marked Terminating
2. Endpoints removed from Service (readiness fails)
3. preStop hook runs (optional sleep)
4. SIGTERM sent to container
5. grace period elapses -> SIGKILL

Align app shutdown timeout with pod terminationGracePeriodSeconds (app timeout < grace period).

Fastify Shutdown

import Fastify from "fastify";
 
const app = Fastify();
app.addHook("onClose", async () => {
  await pool.end();
});
 
const close = async () => {
  await app.close();
  process.exit(0);
};
process.on("SIGTERM", () => void close());

app.close() triggers onClose hooks in reverse registration order.

In-Flight Work Beyond HTTP

Resource	Close API
PostgreSQL `pg.Pool`	`pool.end()`
Redis `ioredis`	`redis.quit()`
BullMQ worker	`worker.close()`
Cron `setInterval`	`clearInterval` + await last job

Track background jobs; server.close() alone does not wait for queue workers.

`@godaddy/terminus`

import { createTerminus } from "@godaddy/terminus";
 
createTerminus(server, {
  signals: ["SIGTERM", "SIGINT"],
  healthChecks: { "/health": () => Promise.resolve() },
  onSignal: async () => { await pool.end(); },
});

Bundles health checks and shutdown for Express apps.

Gotchas

No readiness fail on shutdown - LB sends traffic to dying pod. Fix: shuttingDown flag on /ready.
server.close() without timeout - hang forever on keep-alive. Fix: force exit timer; tune keepAliveTimeout on server.
Closing pool before HTTP drain - in-flight requests error. Fix: server.close() first, then pools.
SIGTERM ignored in dev - Ctrl+C sends SIGINT; handle both.
Workers keep consuming after HTTP stops - duplicate processing during deploy. Fix: pause workers on SIGTERM.
PM2 kill_timeout too short - hard kill mid-request. Fix: 30s kill_timeout - PM2 & systemd.

Alternatives

Alternative	Use When	Don't Use When
Manual SIGTERM handler	Full control, any framework	Want batteries-included health + drain
@godaddy/terminus	Express HTTP services	Fastify (use `app.close()`)
Kubernetes only (no app handler)	Never	Production APIs (always handle SIGTERM)
Immediate `process.exit(0)`	Batch jobs	HTTP services (causes 502s)

FAQs

How long should preStop sleep be?

3-5 seconds often enough for endpoint propagation before SIGTERM. Tune with load balancer docs.

Does Cloud Run send SIGTERM?

Yes on instance scale-in and revision deploy. Same handler pattern applies.

Lambda needs graceful shutdown?

Short-lived invocations; callbackWaitsForEmptyEventLoop matters more than SIGTERM. See Lambda Handler Patterns.

WebSocket connections?

Close WebSocket server in shutdown sequence; notify clients with close frame. May need longer grace period.

NestJS?

app.enableShutdownHooks() listens for SIGTERM and closes modules implementing OnModuleDestroy.

What exit code on forced shutdown?

Exit 1 on timeout so supervisor logs abnormal termination and alerts fire.

Zero-Downtime Deploys - rollout + drain
Health & Readiness Probes - readiness during drain
Kubernetes Deployment - preStop and grace period
Runtime Ops Basics - supervision model
Runtime Ops Best Practices - section checklist

Node.js API Documentation - Complete reference for core modules and APIs.
Node.js Best Practices (GitHub) - Curated production best practices.
DigitalOcean Node.js Tutorials - Production-oriented how-tos.
Node.js Design Patterns, 4th Edition - Production-grade application patterns.
Distributed Systems with Node.js - Enterprise-ready distributed systems.

Stack versions: This page was written for Node.js 24.18.0 (Active LTS), npm 10+, TypeScript 5.6+, Express 5, Fastify 5, and NestJS 11.

Graceful Shutdown

Search Documentation

Graceful Shutdown