Cannot connect to a dockerized Express.js app from another docker container

Question

I have two containers App and Webserver. Webserver is plain nginx:alpine image and App is expressjs app running on port 3030 under ubuntu:focal. I heard that this is a common practise to use separate containers for application and server. So I added proxy_pass http://app:3030/; to nginx config. Something went wrong and I digged into this a bit. To exclude incorrect nginx setup I checked raw curl requests from webserver to app container with no luck. Here is my docker-compose:

version: '3.5'
services:

  webserver:
    image: nginx:alpine
    env_file: .env
    restart: always
    tty: true
    ports:
      - ${NGINX_HOST_HTTP_PORT}:80
      - ${NGINX_HOST_HTTPS_PORT}:443
    volumes:
      - ${APP_CODE_PATH_HOST}:${APP_DIR}
      - ./nginx/nginx.conf:/etc/nginx/nginx.conf
      - ./nginx/sites/:/etc/nginx/conf.d/
      - ./nginx/ssl/:/etc/ssl/
    depends_on:
      - app
    container_name: ${CONTAINER_NAME_PREFIX}-webserver
    networks:
      - app-network

  app:
    env_file: .env
    restart: on-failure
    tty: true
    build:
      dockerfile: ./docker/app/Dockerfile
      args:
        APP_ENV: ${APP_ENV}
        APP_DIR: ${APP_DIR}
        PRODUCT_ID: ${PRODUCT_ID}
      context: ../
    environment:
      - DEBIAN_FRONTEND=noninteractive
    container_name: ${CONTAINER_NAME_PREFIX}-app
    networks:
      - app-network

networks:
  app-network:
    driver: bridge

I can request express from App container CLI:

/ # curl -X GET http://127.0.0.1:3030/multichannel/4034?uid=a6O5iTje8sR2PESbWpAM
{"status": "OK"}

And now from Webserver:

/ # ping app
PING app (172.19.0.2): 56 data bytes
64 bytes from 172.19.0.2: seq=0 ttl=64 time=0.090 ms
/ # curl -X GET http://172.19.0.2:3030/multichannel/4034?uid=a6O5iTje8sR2PESbWpAM
curl: (7) Failed to connect to 172.19.0.2 port 3030 after 0 ms: Connection refused
/ # curl -X GET http://app:3030/multichannel/4034?uid=a6O5iTje8sR2PESbWpAM
curl: (7) Failed to connect to app port 3030 after 0 ms: Connection refused

Nginx log (don't pay attention on IP it is correct, after containers were rebuilt it changed):

2021/10/29 15:40:46 [error] 24#24: *12 connect() failed (111: Connection refused) while connecting to upstream, client: 172.20.0.1, server: api.test, request: "GET /multichannel/4034?uid=a6O5iTje8sR2PESbWpAM HTTP/1.1", upstream: "http://172.20.0.3:3030/multichannel/4034?uid=a6O5iTje8sR2PESbWpAM", host: "api.test"

Of course Nginx conf has:

    location / {
        proxy_pass http://app:3030;
    }

Why was connection refused? Both containers are in the same network

Answer 1

You need to at least expose the 3030 port on the app container to make it available to other containers:

  app:
    expose:
      - "3030"

See the expose docs.

Alternatively, you could publish the port: either specify both ports (HOST:CONTAINER), or just the container port (a random host port will be chosen):

  app:
    ports:
      - "3030"

Answer 2

Finally I found the reason why I can't get connected. When @devatherock mentioned ARTIFACTORY_URL=http://localhost:8081 I checked app code on what host is used and found this:

server.listen(3030, "127.0.0.1", () => {
  logger.debug(`Server is running on ${serverConfig.port} port`);
});

Then I tried "localhost" as host and still got 502. So it is a time to read the docs. Expressjs server.listen works like nodejs server.listen https://nodejs.org/api/net.html#serverlistenport-host-backlog-callback

If host is omitted, the server will accept connections on the unspecified IPv6 address (::) when IPv6 is available, or the unspecified IPv4 address (0.0.0.0) otherwise. In most operating systems, listening to the unspecified IPv6 address (::) may cause the net.Server to also listen on the unspecified IPv4 address (0.0.0.0).

And the answer is to use '0.0.0.0' or just omit host parameter. 127.0.0.1 or localhost accessible only on the local machine because is bounded to the loopback interface.

Answer 3

As I didn't have the express app's code, I tried to simulate the issue with the docker image of another app with a simplified version of the docker-compose file and nginx config. I'm using a ubuntu machine as well. But the call from nginx to the app goes through fine in my case.

docker-compose.yml:

version: '3.5'
services:

  webserver:
    image: nginx:alpine
    restart: always
    tty: true
    ports:
      - 8081:80
    volumes:
      - ./nginx/nginx.conf:/etc/nginx/nginx.conf
    depends_on:
      - app
    container_name: test-webserver
    networks:
      - app-network

  app:
    image: devatherock/artifactory-badge:latest
    environment:
      - ARTIFACTORY_URL=http://localhost:8081
      - ARTIFACTORY_API_KEY=dummy
    restart: on-failure
    tty: true
    container_name: test-app
    networks:
      - app-network
      
networks:
  app-network:
    driver: bridge

nginx.conf:

events {
}
http {
    server {
        listen 80;
        location / {
            proxy_pass http://app:8080;
        }
    }
}

Even I had the same assumption as Robert, that we need to specify the ports section for the app service, but the call from nginx to app worked even without it. When I hit the endpoint http://localhost:8081/health on nginx, it forwards the request to the app at http://app:8080/health and returns the response {"status":"UP"} that the app's health check returns