system/server/info/public request has massive timeout/fails.

[grimkor]

Describe the bug

With docker images from jellyfin and linuxserver on x86:

When loading the application in the browser the https://example.com/system/info/public takes a long time to resolve, leaving the application in a loading state. Eventually the site will be presented and the request will still not have resolved.

This also prevents the Android application from making a connection with the server.

Blocking the URL through Firefox Development Tools proves this endpoint to be the issue as the page will instead load instantaneously.

To Reproduce

1. Launch docker container, example of docker-compose below:

version: "3.4"
services:
  app:
    image: jellyfin/jellyfin:latest
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=Europe/London
      - UMASK_SET=022
    volumes:
      - /path_to/jellyfin/config:/config
      - /path_to/jellyfin/cache:/cache
      - /path_to/jellyfin/files/Music/:/media:ro
      - /path_to/jellyfin/transcoding:/transcode #optional
    ports:
      - 8096:8096
      - 8920:8920
    restart: unless-stopped

2. Open browser developer console and go to network tab
3. Go to http://<URL>:8096/
4. Observe the "public" XHR request

Expected behavior

To resolve the request in a timely manner.

Logs

System (please complete the following information):

• OS: Docker in Debian 9
• Browser: Tested in all browsers, mobile app cannot connect to server
• Jellyfin Version: 10.4.1 (latest)
• Installed Plugins: None
• Reverse proxy: Traefik, tried without traefik and connecting directly to IP address:port and still get the same issue

Additional context

None

[anthonylavado]

Have you tried with the latest version of 10.4.3? The behaviour might be the same, but weMd just like to confirm. It’s strange that this would take so long though, because it’s a local page from the server...

You might have also been hit by the base URL issue - I think if you went to 10.4.0 at all you would be affected.

[grimkor]

Have you tried with the latest version of 10.4.3? The behaviour might be the same, but weMd just like to confirm. It’s strange that this would take so long though, because it’s a local page from the server...

You might have also been hit by the base URL issue - I think if you went to 10.4.0 at all you would be affected.

Thank you for replying and apologies I have been using 10.4.3, I see in my original post I mis-quoted my Jellyfin version.

I've been playing around with it and looking further into other issues for hints and discovered the following:

• I cannot replicate this outside of a Docker stack as a regular container either from the same server or from my local workstation (Ubuntu 18.04).
• If I set the Networking > Bind to local network address to the Docker server's IP address it will instantly fix the issue until reboot where it will be unable to bind to that address.
• Downgrading to a fresh install of 10.3.x did not resolve the issue.

[anthonylavado]

Just as a test, what happens if you set

network_mode: "host" 

[grimkor]

Just as a test, what happens if you set

network_mode: "host" 

I have tried this before; however with docker swarm there is a return message on deploying stating:

Ignoring unsupported options: network_mode

It still deployed but from my testing it did not make any difference.

[anthonylavado]

Hrm okay. We do mention it in our own Docker instructions, but I don’t see it in the LSIO instructions.

Without it, DLNA doesn’t work, and neither does Live TV I believe.

[anthonylavado]

For clarity, loading that endpoint is supposed to show the local address (among some other details), which is used for connecting some clients and things. We’ve run in to issues with it before, and I guess we have to look closer again.

[grimkor]

Hrm okay. We do mention it in our own Docker instructions, but I don’t see it in the LSIO instructions.

Without it, DLNA doesn’t work, and neither does Live TV I believe.

So in that sense if there's no way to get the network_mode to host it will never be 100% in a Docker swarm. I'll do some reading/research to see what that network_mode does and how I might be able to replicate it in a swarm.

For clarity, loading that endpoint is supposed to show the local address (among some other details), which is used for connecting some clients and things. We’ve run in to issues with it before, and I guess we have to look closer again.

Thanks, I thought as much from my quick browse in the server code (totally note a .NET core dev) and would explain why the server information is also blank in the dashboard.

[PrplHaz4]

@grimkor If you care about LiveTV or DLNA, you may want to look into using a macvlan network config in your swarm.

@anthonylavado /system/info/public is loading fine and returns immediately for me on both 10.4.3 (macvlan) and 10.5-dev (--net=host)

[grimkor]

@grimkor If you care about LiveTV or DLNA, you may want to look into using a macvlan network config in your swarm.

I don't have a use for either of those yet but I'll look it up and give it a try and see if it fixes this issue.

[cvium]

Lots of network changes have since been made. Make a new issue if it persists.