Trying and failing to get a Container/s to use a working VPN Container as it's Network

fbeye · Dec 10th 2025

Hey

I think I am missing something. I will post my VPN container and my docker container..maybe you can see something I am missing?

#1 The VPN container does indeed work correctly, I have verified via curl ifconfig.io

#2 The Docker Containers do work, but keep using my real WAN IP

#3 I did create the docker network in question.

Code

VPN

services:
  alpineqbit:
    image: trigus42/qbittorrentvpn
    container_name: alpineqbit
    environment:
      - PUID=1000  #UID of a user on your system
      - PGID=100  #GID of the user group on your system
      - VPN_TYPE=openvpn
      - VPN_USERNAME=[name]
      - VPN_PASSWORD=[password]
      - WEBUI_PASSWORD=[password]
    volumes:
      - /srv/dev-disk-by-uuid-49fe686e-e61f-451b-94f3-f11f9df5625c/Server/downloads:/downloads
      - /srv/dev-disk-by-uuid-49fe686e-e61f-451b-94f3-f11f9df5625c/Server:/server
      - /srv/dev-disk-by-uuid-49fe686e-e61f-451b-94f3-f11f9df5625c/containers/alpineqbit/config:/config
    ports:
      - 8080:8080
    restart: unless-stopped
    cap_add:
      - NET_ADMIN
    sysctls:
      - net.ipv4.conf.all.src_valid_mark=1
      - net.ipv6.conf.all.disable_ipv6=0
    devices:
      - /dev/net/tun
    networks:
      - my-vpn-network

networks:
  my-vpn-network:
    driver: bridge
    driver_opts:
       com.docker.network.mtu: 1400

Display More

Code

Container

services:
  openwebui:
    image: ghcr.io/open-webui/open-webui:main
    # If you have GPU support configured, use the :cuda tag instead:
    # image: ghcr.io/open-webui/open-webui:cuda
    networks:
      - my-vpn-network
    container_name: openwebui
    restart: unless-stopped
    dns: 
      - 1.1.1.1  
    ports:
      - "3000:8080"
    environment:
      # This tells Open WebUI where to find the Ollama service using its service name (ollama)
      OLLAMA_BASE_URL: http://ollama:11434
      USER_AGENT: "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/100.0.4896.75 Safari/537.36"
      # Optional: Set a secret key for security in production environments
      WEBUI_SECRET_KEY: [key]
    volumes:
      # Persists the Open WebUI data (database, settings)
      - open-webui-data:/app/backend/data
    depends_on:
      - ollama

  ollama:
    image: ollama/ollama
    networks:
      - my-vpn-network
    container_name: ollama
    restart: unless-stopped
    # Use the 'deploy' section with 'gpus: all' for Nvidia GPU support
    # Remove this section for CPU-only use
   # deploy:
  #    resources:
    #    reservations:
    #      devices:
   #         - driver: nvidia
      #        count: all
   #           capabilities: [gpu]
    volumes:
      # Persists the models you download
      - ollama-data:/root/.ollama
    ports:
      - "11434:11434" # Exposes the Ollama API port
    dns:
      - 1.1.1.1  

# Define the volumes for persistence
volumes:
  open-webui-data:
  ollama-data:

networks:
  my-vpn-network:
    external: true

Display More

gderf · Dec 10th 2025

I think that you should not have port statements in containers that use another container's network. They should be defined in the container whose network you want to bind to.

fbeye · Dec 10th 2025

Ahh I actually do recall having to do that in the past. I will try that tomorrow and see how it goes. Although, would that also cause its IP address to be my WAN IP? Anyway, I’ll check it tomorrow, ty.

gderf · Dec 10th 2025

I can't say for sure but it would seem to me that if a container can't join the specified network of another container then the current system local network would be used and that would route out via your WAN IP.

Are you referring to any specific guides to do what you are trying to accomplish, there are plenty of examples out there. ChatGPT is pretty good about things like this.

fbeye · Dec 10th 2025

This particular container, llama and open-we up are just for fun. We can be talking about any Container TBH.. I know the qbit/vpn works as should when I curl ifconfig.io and then the stack with web/llama works on its own, I just can’t get the stack to use the vpn for its wan access. Essentially anything being searched or looked up would be behind the vpn.

No guide necessarily, just figured it should be an easy thing.

gderf · Dec 10th 2025

You need to find a guide that specifically tells you how to setup a container to use the network of another container. There are many available. Start looking.

BernH · Dec 10th 2025

Basic guide.

Don't define a network or ports in the container you want to put in the vpn network. Define it's ports in the vpn container.

In the container that you want to put in the vpn use one of these network_mode lines to place it on the network of the vpn container.

network_mode: container:<container_name_or_id>

or

network_mode: service:<service_name>

Here is an example for concept (this uses the service version. container_name must be specified to use the container version):

Code

services:
  web:
    image: nginx

  app:
    image: my-app
    network_mode: service:web # Example: sharing network with the 'web' service

Participate now!