SOLVED: Local Network Access through Proxy

[contabs]

So for my setup, I have Jellyfin and Nginx Proxy Manager in a user-defined docker bridge on an Ubuntu 24.04 server. I have the port-forwarding setup so that I can reach from my domain name when outside the LAN, but when I am on the LAN the domain I cannot access it through the domain name through any device other than my desktop (which buffers heavily), nor the IP address because of the user-defined bridge network. I tried setting up a MACVLAN network through docker but my WAP won't support it. Is there a way that I can access my server internally? All suggestions are welcome, including switching proxies/hosting options if necessary, I'm open to anything.

Edit: I do have control of my local DNS server through AdGuard so I can try a DNS entry, but only once I can access it from the IP locally.

[TheDreadPirate]

I'm assuming that Nginx Proxy Manger is using host networking or using "port" to publish ports. And your jellyfin container is using "expose". Is that a correct assumption?

Or are both using "ports" to publish their ports to the host interface?

[contabs]

Looking at my docker-config for jellyfin, I didn't specify a port, however it does it automatically is how it got 8096 assigned. When I run 'docker ps', it shows that the jellyfin container has 8096 attached to it, and it specifies 8096 while inspecting the container. For NPM, I did specify "ports" in the compose file.

[TheDreadPirate]

So the container is using port 8096, but that is not pushed to the host's interface without the "ports" section in your compose. Without "ports", only other containers on the same bridge will be able to access Jellyfin, which is OK if you are using NPM for all connections and NPM is on the same bridge network.

Having said that, I do believe you at least need "expose" on your jellyfin container so that NPM can connect on that port.

Code:

jellyfin-testing:
    image: jellyfin/jellyfin:10.9.11
    container_name: jellyfin-testing
    hostname: jellyfin-testing
    user: 1002:1002
    group_add:
      - "993" # Intel render group
    networks:
      blackmoon:
        ipv4_address: 172.16.100.2
    expose:
      - 8096
    volumes:
      - ./jellyfin-stable-data/config:/config
      - ./jellyfin-stable-data/cache:/cache
      - /media/library:/media/library:ro
      - /media/storage2:/media/storage2
      - /media/ramdisk:/media/ramdisk
    devices:
      - /dev/dri/renderD128:/dev/dri/renderD128
    environment:
      - TZ=America/New_York
    restart: unless-stopped

[contabs]

Okay so after a few attempts with expose it worked with 'ports'. For anyone in the future who has this issue, NPM is still forwarding to 8096, even though it assigned me a random ephemeral port bc I forgot to specify lol. Here is my redacted docker-compose.yml:

services:
jellyfin:
image: jellyfin/jellyfin
container_name: jellyfin
user: 1000:1000
network_mode: '<user defined bridge name>'
ports:
- 8096 # Unless you specify with port:port it will assign a random port that forwards to 8096
volumes:
- /path/to/configconfig
- /path/to/cachecache
- /path/to/mediamedia
restart: 'unless-stopped'

Thanks for the help!