Jellyfin GPU transcoding within Docker

#1

Does anyone have any information about how to use your host’s GPU (without taking it away from the host) so that Jellyfin can access it for hardware transcoding? I know this is more of a Docker question, but there isn’t much clear information out there, and I thought someone here would have already done it.

The host is Ubuntu 18.04.2, Jellyfin is running in Docker, and the video card is a Quatro P2000. The host already has the P2000 drivers installed and verified working … I’d just like the Docker for Jellyfin to be able to utilize it without reserving it just for the Docker image.

#2

Not sure if this helps but there are some detailed instructions for the unraid Nvidia plugin and docker. Jellyfin/Emby should be the same instructions wise.

Maybe some of the instructions are similar to help you on the way?

#3

Thanks … I decided that splitting the card between two loaded sets of graphics drivers was probably not the best idea for stability, and moved my installation to the Ubuntu PPA.

If I could make a suggestion to the devs for a nice improvement, an export of the config and metadata that allows import later would be amazing. I moved/chown’d/set permissions on everything, and I had a bunch of broken functionality after moving. In the end, I had to set the server up from scratch again and send everyone password resets.