x-forwarded-for docker-compose network

[La-source]

Hello,

By setting nginx-proxy with docker-compose and an external network the remote ip address is the internal IP address and not the client IP address.

version: '2'

services:
  app:
    image: node:12-alpine
    restart: always
    command: sh -c "npm install && npm start"
    working_dir: /app
    expose:
      - 3000
    volumes:
      - ./dist:/app
    environment:
      - VIRTUAL_HOST=domain.com
      - VIRTUAL_PORT=3000
  proxy:
    image: jwilder/nginx-proxy
    restart: always
    ports:
      - "80:80"
      - "443:443"
    volumes:
      - ./vhost.d:/etc/nginx/vhost.d
      - ./certs:/etc/nginx/certs:ro
      - ./html:/usr/share/nginx/html
      - /var/run/docker.sock:/tmp/docker.sock:ro
    environment:
      - ENABLE_IPV6=true
    logging:
      driver: "none"

networks:
  default:
    external:
      name: nginx-proxy

Result:

'x-real-ip': '172.23.0.1',
'x-forwarded-for': '172.23.0.1',
'x-forwarded-proto': 'https',
'x-forwarded-ssl': 'on',
'x-forwarded-port': '443',

[La-source]

Hello,

After a lot of research I noticed that it was a real problem with docker and that help on this subject is quite poor.

Concretely, the solution that worked for me is to add network_mode: 'host' in nginx-proxy to get the correct ip address.

In the case of a docker-dial version 3 it is necessary to follow this syntax: moby/moby#25526 (comment)

I think an addition to the documentation will not be too much.

[Munzy]

Can you show an example of your config?

[windmemory]

This issue is related to docker/roadmap#157

I am also experiencing this problem, and seems like the network_mode: host is the only solution now.

[rjhancock]

network_mode: host only works on Linux and is not suitable for any other platform.