Resilience Basics

8 examples to get you started with Resilience for Node.js backends - 6 basic and 2 intermediate.

Prerequisites

• Node.js 24.18.0 with native fetch and AbortSignal.timeout.
• HTTP service behind Kubernetes, ECS, or similar for health/readiness probes.

Basic Examples

1. Timeout Outbound fetch

Hung downstream calls exhaust your connection pool.

const response = await fetch("https://api.partner.com/data", {
  signal: AbortSignal.timeout(5_000),
});

• AbortSignal.timeout available in Node 18+ - no manual setTimeout.
• Map TimeoutError to 504 or retry policy upstream.
• Shorter timeouts on user-facing paths than batch jobs.

Related: Timeouts Everywhere - full timeout matrix

2. Fail Fast on Boot Config

Do not serve traffic with invalid environment.

import { z } from "zod";
 
const Env = z.object({
  DATABASE_URL: z.string().url(),
  PORT: z.coerce.number().default(3000),
});
 
export const env = Env.parse

• Process exits non-zero - orchestrator restarts with fixed config.
• Readiness stays false until DB migration completes if you gate on that.
• Better one failed deploy than silent runtime errors on every request.

Related: Configuration Basics - Zod env parse

3. Liveness vs Readiness

Tell Kubernetes whether to restart vs stop sending traffic.

app.get("/health", (_req, res) => res.json({ ok: true }));
 
app.get("/ready", async (_req, res) => {
  try {
    await

• Liveness: process not deadlocked - cheap check.
• Readiness: dependencies available - fails when DB down so kube removes from LB.
• Do not put slow checks on liveness - restart loops.

Related: Observability Basics - probe patterns

4. Graceful Shutdown on SIGTERM

Drain in-flight HTTP before exit on deploy.

const server = app.listen(3000);
 
process.on("SIGTERM", () => {
  server.close(() => {
    logger.info({ event: "shutdown_complete" });
    process.exit(0);

• server.close stops accepting new connections; finishes active ones.
• Force exit after 30s if drain stalls - match k8s terminationGracePeriodSeconds.
• Close DB pools and OTel SDK in same shutdown handler.

Related: Graceful Degradation - feature-off modes

5. Retry Only When Safe

POST without idempotency keys can double-charge.

async function safeGetWithRetry(url: string, attempts = 3): Promise<Response> {
  for (let i = 0; i < attempts; i++) {
    try {
      const

• Retry GET and idempotent PUT - safe by HTTP semantics (with caveats).
• Retry 5xx, not 4xx (client errors will not heal).
• Exponential backoff reduces thundering herd.

Related: Retries with Backoff - jitter and idempotency

6. Return Explicit Errors, Not Hangs

Users and load balancers need a response within SLA.

app.use((req, res, next) => {
  const timer = setTimeout(() => {
    if (!res.headersSent) {
      res.status(504).json({ error: "Gateway Timeout" });

• Server-level timeout is last resort - prefer per-outbound timeouts.
• Log 504 with requestId for correlation.
• Align with LB idle timeout (often 60s).

Related: Timeouts Everywhere - server timeouts

Intermediate Examples

7. Circuit Breaker Sketch

Stop calling a failing dependency to let it recover.

let failures = 0;
const THRESHOLD = 5;
let openUntil = 0;
 
async function callBilling() {
  if (Date.now() < openUntil) throw new Error("Circuit open"

• Open circuit fast-fails local requests - saves threads and downstream load.
• Half-open retry after cooldown - production use opossum library.
• Pair with graceful degradation for user experience.

Related: Circuit Breakers - opossum patterns

8. Bulkhead: Limit Concurrent Outbound Calls

One slow vendor cannot consume all sockets.

import { Semaphore } from "async-mutex"; // or custom pool
 
const vendorSem = new Semaphore(10);
 
async function callVendor() {
  const [, release] = await vendorSem.acquire();
  try {

• Cap concurrent calls to fragile integration - bulkhead pattern.
• Remaining requests fail fast locally instead of queuing forever.
• Tune limit from vendor SLA and your pool size.

Related: Circuit Breakers - bulkhead with opossum

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.