Server Version#: 1.41.4.9463
Player Version#: NA
I’ve migrated my Plex docker container (plexinc/pms-docker:plexpass) from an older Intel i7-4700K Ubuntu 24.04 desktop to an unprivileged LXC running on Proxmox 8.3.3 (Intel i7-8700, 6.8.12-8-pve). The LXC is Debian Bookworm 12 with all recent updates as of Feb 16, 2025. I’ve given the LXC 4 CPUs, 4GB of RAM, and 60GB of disk space. I’ve also passed through my Intel A380 GPU using the new features in PVE 8.3 where you only need to share the /dev/dri/card# and /dev/dri/renderD12#. On the LXC, I’m running Docker. In my compose.yaml file, I’ve got the following:
devices:
- /dev/dri:/dev/dri
# - /dev/dri/card3:/dev/dri/card3
# - /dev/dri/renderD129:/dev/dri/renderD129
You can see I’ve tried the individual card and renderD names, but they produced the same results. So I went with what is officially documented here. My issue is that Plex will not utilize the Arc380 for transcoding at all. It detects it in the TRANSCODING settings page as INTEL DG2 [Arc A380]. What’s strange is that if I go into the container, I can see the following permissions/groups on /dev/dri:
root@plex:/dev/dri# ls -lha
total 0
drwxr-xr-x 2 root root 80 Feb 17 16:30 .
drwxr-xr-x 6 root root 360 Feb 17 16:30 ..
crw-rw---- 1 root video 226, 3 Feb 17 16:30 card3
crw-rw---- 1 root 104 226, 129 Feb 17 16:30 renderD129
root@plex:/dev/dri#
You can see here that the container has no “render” group and is listing the GID for the LXC container. I don’t even see a render group in the container at all, only a video group:
root@plex:/# cat /etc/group | grep render
root@plex:/# cat /etc/group | grep video
video:x:44:
root@plex:/#
If I chmod 666 /dev/dri/renderD129 then the transcoding operation works as expected. If I reboot the LXC container, I lose that chmod setting and I have to manually change it again. What am I missing here? Should the renderD129 node be a member of video?? Is that the default? If so, what do I need to do to change this? Thanks for your help!