Production Deployment

Problem

You need to deploy an ergo API to production with process management, a reverse proxy for TLS termination, health checks for load balancer integration, structured logging for observability, Redis-backed rate limiting for multi-instance consistency, and graceful shutdown for zero-downtime deploys. These concerns are documented individually across several recipe and middleware pages — you need a consolidated reference showing how they compose into a complete deployment.

Solution

Application Setup

Wire environment-based configuration, structured logging, Redis rate limiting, a health check endpoint, and graceful shutdown into a single entry point:

Standalone

import pino from 'pino';
import Redis from 'ioredis';
import {
  compose,
  handler,
  logger,
  accepts,
  url,
  rateLimit,
} from '@centralping/ergo';
import {graceful} from '@centralping/ergo-router';

const {
  PORT = '3000',
  HOSTNAME = '0.0.0.0',
  REDIS_URL = 'redis://localhost:6379',
  SHUTDOWN_TIMEOUT = '10000',
} = process.env;

const log = pino({level: 'info'});
const redis = new Redis(REDIS_URL);

class RedisStore {
  constructor(client) {
    this.redis = client;
  }

  async hit(key, windowMs) {
    const now = Date.now();
    const windowKey = `rl:${key}`;
    const cutoff = now - windowMs;

    const pipeline = this.redis.pipeline();
    pipeline.zremrangebyscore(windowKey, 0, cutoff);
    pipeline.zadd(windowKey, now, `${now}:${Math.random()}`);
    pipeline.zcard(windowKey);
    pipeline.zrange(windowKey, 0, 0, 'WITHSCORES');
    pipeline.pexpire(windowKey, windowMs);

    const results = await pipeline.exec();
    const count = results[2][1];
    const oldest = results[3][1];
    const oldestScore = oldest.length >= 2 ? Number(oldest[1]) : now;
    const resetMs = Math.max(0, oldestScore + windowMs - now);

    return {count, resetMs};
  }
}

const proxyKeyGenerator = (req) => {
  const forwarded = req.headers['x-forwarded-for'];
  return forwarded
    ? forwarded.split(',')[0].trim()
    : req.socket.remoteAddress;
};

let isShuttingDown = false;

const healthPipeline = handler(
  (req, res, acc) => {
    if (isShuttingDown) {
      return {response: {statusCode: 503, body: {status: 'draining'}}};
    }
    return {response: {body: {status: 'healthy'}}};
  },
);

const apiPipeline = handler(
  compose(
    logger({log: (info) => log.info(info)}),
    accepts({types: ['application/json']}),
    url(),
    rateLimit({
      max: 100,
      windowMs: 60_000,
      store: new RedisStore(redis),
      keyGenerator: proxyKeyGenerator,
    }),
    (req, res, acc) => ({
      response: {body: {path: acc.url.pathname}},
    }),
  ),
);

const requestListener = (req, res) => {
  if (req.url === '/health') return healthPipeline(req, res);
  return apiPipeline(req, res);
};

const {server, shutdown} = await graceful(requestListener, {
  port: Number(PORT),
  hostname: HOSTNAME,
  timeout: Number(SHUTDOWN_TIMEOUT),
  log,
  onStartup: async ({log}) => {
    await redis.ping();
    log.info('Redis connected');
  },
  onShutdown: async ({log}) => {
    isShuttingDown = true;
    await redis.quit();
    log.info('Redis disconnected');
  },
});

ergo-router

import pino from 'pino';
import Redis from 'ioredis';
import createRouter, {graceful} from '@centralping/ergo-router';

const {
  PORT = '3000',
  HOSTNAME = '0.0.0.0',
  REDIS_URL = 'redis://localhost:6379',
  SHUTDOWN_TIMEOUT = '10000',
} = process.env;

const log = pino({level: 'info'});
const redis = new Redis(REDIS_URL);

class RedisStore {
  constructor(client) {
    this.redis = client;
  }

  async hit(key, windowMs) {
    const now = Date.now();
    const windowKey = `rl:${key}`;
    const cutoff = now - windowMs;

    const pipeline = this.redis.pipeline();
    pipeline.zremrangebyscore(windowKey, 0, cutoff);
    pipeline.zadd(windowKey, now, `${now}:${Math.random()}`);
    pipeline.zcard(windowKey);
    pipeline.zrange(windowKey, 0, 0, 'WITHSCORES');
    pipeline.pexpire(windowKey, windowMs);

    const results = await pipeline.exec();
    const count = results[2][1];
    const oldest = results[3][1];
    const oldestScore = oldest.length >= 2 ? Number(oldest[1]) : now;
    const resetMs = Math.max(0, oldestScore + windowMs - now);

    return {count, resetMs};
  }
}

const proxyKeyGenerator = (req) => {
  const forwarded = req.headers['x-forwarded-for'];
  return forwarded
    ? forwarded.split(',')[0].trim()
    : req.socket.remoteAddress;
};

let isShuttingDown = false;

const router = createRouter({
  transport: {
    trustProxy: true,
  },
  defaults: {
    accepts: {types: ['application/json']},
    logger: {
      log: (info) => log.info(info),
      error: (info) => log.error(info),
    },
    rateLimit: {
      max: 100,
      windowMs: 60_000,
      store: new RedisStore(redis),
      keyGenerator: proxyKeyGenerator,
    },
  },
});

router.get('/health', {
  authorization: false,
  rateLimit: false,
  execute: () => {
    if (isShuttingDown) {
      return {response: {statusCode: 503, body: {status: 'draining'}}};
    }
    return {response: {body: {status: 'healthy'}}};
  },
});

router.get('/users', {
  authorization: {
    strategies: [
      {
        type: 'Bearer',
        authorizer: async (attributes, token) => {
          if (!token) return {authorized: false};
          return {authorized: true, info: {userId: token}};
        },
      },
    ],
  },
  execute: () => ({
    response: {body: [{id: '1', name: 'Jane'}]},
  }),
});

await graceful(router.handle(), {
  port: Number(PORT),
  hostname: HOSTNAME,
  timeout: Number(SHUTDOWN_TIMEOUT),
  log,
  onStartup: async ({log}) => {
    await redis.ping();
    log.info('Redis connected');
  },
  onShutdown: async ({log}) => {
    isShuttingDown = true;
    await redis.quit();
    log.info('Redis disconnected');
  },
});

Tip

The RedisStore implementation above uses the same ioredis sorted-set pattern documented in the rate-limit middleware guide. See that page for the full store contract and when to use a shared store.

trustProxy requires a trusted reverse proxy

Only enable trustProxy: true when the application runs behind a reverse proxy you control (nginx, a cloud load balancer, etc.). Without a trusted proxy, clients can forge X-Forwarded-* headers. See the ergo-router transport options for details.

Proxy-aware rate limiting

The code examples above include a proxyKeyGenerator that reads X-Forwarded-For instead of req.socket.remoteAddress. Without this, all clients behind a reverse proxy share a single rate-limit bucket. See the rate-limit middleware guide — Proxy Awareness for details and security caveats.

Dockerfile

Use a multi-stage build to keep the production image minimal:

FROM node:22-alpine AS deps
WORKDIR /app
COPY package.json package-lock.json ./
RUN npm ci --omit=dev

FROM node:22-alpine
RUN apk add --no-cache tini
WORKDIR /app
ENV NODE_ENV=production
COPY --from=deps /app/node_modules ./node_modules
COPY . .
USER node
EXPOSE 3000
ENTRYPOINT ["tini", "--"]
CMD ["node", "server.js"]

Tip

tini acts as PID 1 and forwards signals correctly to the Node.js process. Without an init process, SIGTERM from docker stop may not reach your application. An alternative is init: true in your docker-compose.yml service definition (see below).

docker-compose.yml

Compose the application with nginx and Redis:

services:
  app:
    build: .
    init: true
    environment:
      - PORT=3000
      - HOSTNAME=0.0.0.0
      - REDIS_URL=redis://redis:6379
      - SHUTDOWN_TIMEOUT=10000
    depends_on:
      redis:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "wget", "-qO-", "http://localhost:3000/health"]
      interval: 10s
      timeout: 5s
      retries: 3
      start_period: 5s

  nginx:
    image: nginx:alpine
    ports:
      - "443:443"
      - "80:80"
    volumes:
      - ./nginx.conf:/etc/nginx/nginx.conf:ro
      - ./certs:/etc/nginx/certs:ro
    depends_on:
      app:
        condition: service_healthy

  redis:
    image: redis:7-alpine
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 10s
      timeout: 5s
      retries: 3

nginx Configuration

Configure nginx as a reverse proxy with TLS termination:

worker_processes auto;

events {
    worker_connections 1024;
}

http {
    upstream api {
        server app:3000;
    }

    server {
        listen 80;
        return 301 https://$host$request_uri;
    }

    server {
        listen 443 ssl;

        ssl_certificate     /etc/nginx/certs/cert.pem;
        ssl_certificate_key /etc/nginx/certs/key.pem;
        ssl_protocols       TLSv1.2 TLSv1.3;

        location / {
            proxy_pass http://api;

            proxy_set_header Host              $host;
            proxy_set_header X-Real-IP         $remote_addr;
            proxy_set_header X-Forwarded-For   $proxy_add_x_forwarded_for;
            proxy_set_header X-Forwarded-Proto $scheme;

            proxy_read_timeout 60s;
            proxy_send_timeout 60s;
        }
    }
}

Explanation

Trust Proxy Configuration

When running behind a reverse proxy, enable trustProxy: true on the router so that ergo-router reads client identity from X-Forwarded-For instead of the direct socket address, and X-Forwarded-Proto for protocol detection. This affects:

• Request IDs — the request-id transport reads X-Request-Id / X-Forwarded-For headers when trusted
• HSTS — security headers use X-Forwarded-Proto to determine whether to send Strict-Transport-Security

Never trust proxies from the public internet

If your application is directly exposed to the internet (no reverse proxy), leave trustProxy disabled. A malicious client can set X-Forwarded-For to any value, spoofing their IP address for rate limiting, logging, and access control.

Environment-Based Configuration

The application reads all deployment-specific values from environment variables with sensible defaults:

Variable	Default	Purpose
PORT	3000	HTTP listen port
HOSTNAME	0.0.0.0	Bind address
REDIS_URL	redis://localhost:6379	Redis connection string
SHUTDOWN_TIMEOUT	10000	Graceful shutdown deadline (ms)

This keeps the application image immutable — the same image runs in development, staging, and production with different environment values.

Connection Draining

The timeout option on graceful() sets the maximum time to wait for in-flight requests to complete after a shutdown signal. Align this value with your load balancer’s drain timeout:

docker stop --time=30 app    ← sends SIGTERM, waits 30s
graceful(..., { timeout: 25000 })  ← app drains for 25s, then force-closes

Set the application timeout slightly below the container stop timeout to ensure the process exits cleanly before the container runtime force-kills it. See the Graceful Shutdown recipe for the complete lifecycle flow and configuration options.

TLS Termination

nginx handles TLS termination — the application receives plain HTTP on its internal port. This separates certificate management from application code and allows certificate rotation without application restarts.

The X-Forwarded-Proto: https header from nginx tells ergo-router (via trustProxy: true) that the original client connection was encrypted, enabling correct HSTS header generation and protocol-aware redirects.

Process Signals in Containers

Docker sends SIGTERM to PID 1 when stopping a container. Node.js processes that are PID 1 do not handle SIGTERM by default — use one of these approaches:

Approach	How	Trade-off
tini (Dockerfile ENTRYPOINT)	ENTRYPOINT ["tini", "--"]	Adds ~20KB to image; handles zombie reaping
init: true (docker-compose)	init: true in service definition	No image change; Docker provides the init process
--init (docker run)	docker run --init ...	Per-invocation flag

graceful() listens for SIGINT and SIGTERM by default and initiates the shutdown sequence automatically. The init process ensures the signal reaches the Node.js process.

What This Recipe Does Not Cover

• Kubernetes — orchestration, pod lifecycle, and readiness/liveness probes are out of scope. The health check endpoint works with any orchestrator.
• CI/CD pipelines — image building and deployment automation depend on your platform.
• Secrets management — environment variables are shown for simplicity. Use your platform’s secret management (Docker secrets, Vault, cloud KMS) for production credentials.
• Monitoring and alerting — see Structured Logging for log integration patterns and the Graceful Shutdown recipe for OpenTelemetry SDK lifecycle.