+- Jellyfin Forum (https://forum.jellyfin.org)
+-- Forum: Off Topic (https://forum.jellyfin.org/f-off-topic)
+--- Forum: Self-hosting & Homelabs (https://forum.jellyfin.org/f-self-hosting-homelabs)
+--- Thread: Docker vs normal install (/t-docker-vs-normal-install)
Docker vs normal install - Azriall - 2024-04-15
Good morning
I'm currently trying to fix my Jellyfin as we speak. Can't access the webui. But this has got me thinking. I see allot of posts on the web about installing Jellyfin on docker. Why is this the way to go? if so How i would make a folder with media on it that It can read? Still learning about docker and its uses. Coming from a new to ubuntu perspective but not a total noob. I'm currently using a pc with ubuntu 22.04 Server LTS. This is my personal home project that I was given a old PC to make use and seeing what I can do with it.
RE: Docker vs normal install - bitmap - 2024-04-15
For me? Docker fits into my ecosystem and provides a small bit of additional isolation of the program from the rest of the system. Additionally, if you screw something up with a container, you can start fresh super easily without much (if any) data loss. Plus, if you upgrade hardware, migration is a breeze compared to bare metal install and you never have to worry about dependencies (caveat being proper setup and drivers for hardware acceleration).
If you're ONLY running Jellyfin, you're probably fine with either route. If you want to learn docker/compose, go for it. If you're running a whole slew of services, docker is the way to go, compose specifically.
RE: Docker vs normal install - Azriall - 2024-04-15
I would like to run a few programs, Samba for file sahring, something to monitor net useage or performace, jellyfin and a version of ubuntu I could access on the web as a private service. And if it can be done my own AI model via ollama. As well as learn more about networking on linux and home servers
RE: Docker vs normal install - bitmap - 2024-04-15
Most of that sounds like bare metal. Docker will have you learn more about Docker networking than Linux networking, but it requires a small amount of knowledge in both spaces. Samba isn't really a containerized application, nor is performance monitoring (though it is certainly an option). Only way I found to easily run an AI model in Docker is via Nextcloud, which I had set up successfully prior to the most recent upgrade with both GPT-4 and Stable Diffusion, although you can absolutely use llama if you'd like. There is not a clean way to run a VM of Ubuntu in a Docker container, although there are some interesting workarounds out there. You'd be better off with remote access via VPN.
Sounds like bare metal might be the way to go at the moment since most of your aims align better with that. If you ever decide to run a bunch of things (e.g., Wireguard node, VPN for services/containers, Servarr suite, Nextcloud, Matrix server), then I'd say you should look at docker-compose. For now, sounds like you may want to keep it simple unless you find something that steers you towards Docker.