Jellyfin hardware acceleration docker issues.
from hoxbug@lemmy.world to selfhosted@lemmy.world on 21 Sep 08:49
https://lemmy.world/post/20015319

Having a bit of trouble getting hardware acceleration working on my home server. The cpu of the server is an i7-10700 and has a discrete GPU, RTX 2060. I was hoping to use intel quick sync for the hardware acceleration, but not having much luck.

From the guide on the jellyfin site jellyfin.org/docs/general/administration/…/intel

I have gotten the render group ID using “getent group render | cut -d: -f3” though it mentions on some systems it might not be render, it may be video or input which i tried with those group ID’s as well.

When I run “docker exec -it jellyfin /usr/lib/jellyfin-ffmpeg/vainfo” I get back

libva info: VA-API version 1.22.0
libva info: Trying to open /usr/lib/jellyfin-ffmpeg/lib/dri/nvidia_drv_video.so
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/nvidia_drv_video.so
libva info: Trying to open /usr/lib/dri/nvidia_drv_video.so
libva info: Trying to open /usr/local/lib/dri/nvidia_drv_video.so
libva info: va_openDriver() returns -1
vaInitialize failed with error code -1 (unknown libva error),exit

I feel like I need to do something on the host system since its trying to use the discrete card? But I am unsure.

This is the compose file just in case I am missing something

version: "3.8"
services:
  jellyfin:
    image: jellyfin/jellyfin
    user: 1000:1000
    ports:
      - 8096:8096
    group_add:
      - "989" # Change this to match your "render" host group id and remove this comment
      - "985"
      - "994"
    # network_mode: 'host'
    volumes:
      - /home/hoxbug/Docker/jellyfin/config:/config
      - /home/hoxbug/Docker/jellyfin/cache:/cache
      - /mnt/External/Movies:/Movies
    devices:
      - /dev/dri/renderD128:/dev/dri/renderD128
networks:
  external:
    external: true

Thank you for the help.

#selfhosted

threaded - newest

sk@hub.utsukta.org on 21 Sep 09:03 next collapse

On my system i was able to use my integrated iGPU wit the follwing:

    devices:
     - /dev/dri:/dev/dri

The rest of your compose looks fine.
sneezycat@sopuli.xyz on 21 Sep 09:59 next collapse

Isn’t your GPU an Nvidia RTX 2060? Why are you trying to use the Intel GPU acceleration method? I’m confused

hoxbug@lemmy.world on 21 Sep 10:47 collapse

It just seemed the easiest route, but I may just give using the GPU a go.

__ghost__@lemmy.ml on 21 Sep 13:57 collapse

From personal experience intel QSV wasn’t worth the trouble to txshoot on my hardware. Mine is a lot older than yours though. Vaapi has worked well on my arc card

entropicdrift@lemmy.sdf.org on 21 Sep 16:59 collapse

QSV is the highest quality video transcoding hardware acceleration out there. It’s worth using if you have a modern Intel CPU (8th gen or newer)

bobslaede@feddit.dk on 21 Sep 11:05 next collapse

This is how mine works, with a Nvidia GPU

services:
  jellyfin:
    volumes:
      - jellyfin_config:/config
      - jellyfin_cache:/cache
      - type: tmpfs
        target: /cache/transcodes
        tmpfs:
          size: 8G
      - media:/media
    image: jellyfin/jellyfin:latest
    restart: unless-stopped
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids:
                - "0"
              capabilities:
                - gpu
notfromhere@lemmy.ml on 21 Sep 12:52 collapse

What is the tmpfs for?

bobslaede@feddit.dk on 21 Sep 13:14 collapse

Temp files for transcoding. No need to hit the disk.

__ghost__@lemmy.ml on 21 Sep 13:56 collapse

Is the tmpfs on RAM?

baduhai@sopuli.xyz on 21 Sep 14:26 next collapse

Aye.

bobslaede@feddit.dk on 21 Sep 18:44 next collapse

Hmm. I would think so. But I haven’t actually checked. That was my thought.

486@lemmy.world on 22 Sep 08:41 collapse

It is. It might end up on disk in swap, if you run low on memory (and have some sort of disk-based swap enabled), but usually it is located in RAM.

__ghost__@lemmy.ml on 22 Sep 17:47 collapse

You can create a tmpfs on other storage devices as well, just curious what their setup looked like

486@lemmy.world on 24 Sep 09:33 collapse

No, tmpfs is always located in virtual memory. Have a look at the kernel documentation for more information about tmpfs.

HumanPerson@sh.itjust.works on 21 Sep 15:33 next collapse

I have an arc for transcoding, and I had to set the device to /dev/dri without the renderD128 part. If I were you, I would just use the 2060. If it’s there for llama or something I’d still try it and see how it does doing both at once, as it should be separate parts of the gpu handling that.

entropicdrift@lemmy.sdf.org on 21 Sep 17:01 next collapse

If you switch the devices line to

- /dev/dri:/dev/dri

as other have suggested, that should expose the Intel iGPU to your Jellyfin docker container. Presently you’re only exposing the Nvidia GPU.

JustEnoughDucks@feddit.nl on 22 Sep 07:18 collapse

Do you have the Intel drivers installed on your machine? Are GuC and HuC working?

sudo reboot
sudo dmesg | grep i915
sudo cat /sys/kernel/debug/dri/0/gt/uc/guc_info
sudo cat /sys/kernel/debug/dri/0/gt/uc/huc_info

On Debian I had to manually download the i915 full driver Zip, extract it, take out the Intel drivers, and put it in /usr/lib/firmware

Then hardware acceleration worked on my Arc380.

If you use QSV, your CPU iGPU will be the one that can use it, so make sure to set your render device in docker to the iGPU and not the RTX 2060