SOLVED: Hardware Transcoding Issue

[SNChalmers]

Hey folks,

Hardware transcoding used to work fine, and sometime over the past couple months it just stopped working. The issue appears when I try to play HEVC media in a browser, Jellyfin says "Playback Error Playback failed due to a fatal player error." This has happened before, but the issue was that I had run out of storage, which is not the case this time. Thanks in advance for any help

My server is running Debian 11 and OpenMediaVault. Jellyfin is running in Docker. and is version 10.9.7. The CPU is a Ryzen 1600 with reasonably up to date firmware, 16GB of RAM, and a GTX1060. I have attached logs from the last time I attempted to play an HEVC file. Please let me know what other information I can provide that would be helpful.

Thanks again!

Logs: https://pastebin.com/1eC7sn8p

[TheDreadPirate]

We'd need to see the ffmpeg log, located in the same directory as the jellyfin log.

Since you are on Debian 11, I am betting that your Nvidia driver and/or container toolkit is old.

[SNChalmers]

Here are the three ffmpeg logs that appear to have been generated at the same time as the log previously posted.

https://pastebin.com/0xbq49iQ
https://pastebin.com/nqzLhxet
https://pastebin.com/KVkui5Ze

[SNChalmers]

The relevant lines appear to be
[AVHWDeviceContext @ 0x5611a6dbaf00] Cannot load libcuda.so.1
[AVHWDeviceContext @ 0x5611a6dbaf00] Could not dynamically load CUDA
Device creation failed: -1.
Failed to set value 'cuda=cu:0' for option 'init_hw_device': Operation not permitted
Error parsing global options: Operation not permitted

But after some googling I'm still unsure how to fix this.

[TheDreadPirate]

Are you using docker compose to configure the container or OMV's UI?

[SNChalmers]

Docker compose, through Portainer.

[TheDreadPirate]

Can you share your docker compose config?

[SNChalmers]

Apologies for not including it initially, somehow my password to Portainer got reset and I was struggling to figure out how to log in lol. Jellyfin specifically is this;

EDIT: Not sure what happened to the formatting here but in Portainer everything is indented correctly, and obviously using colon+slash instead of lol

version: "3"
services:
  jellyfin:
  # image: linuxserver/jellyfin:10.8.11-1-ls229
    image: ghcr.io/jellyfin/jellyfin:latest
  #  image: jellyfin/jellyfin
    container_name: jellyfin
    environment:
      - PUID=1000
      - PGID=100
      - TZ=America/Vancouver
    #  - UMASK_SET=022 #optional
    volumes:
      - /var/lib/docker/configs/jellyfinconfig
      - /srv/dev-disk-by-uuid-43c6f622-4118-429f-89b7-f9cd3995ea21/RaidSharedBase/Media/TVdata/tvshows
      - /srv/dev-disk-by-uuid-43c6f622-4118-429f-89b7-f9cd3995ea21/RaidSharedBase/Media/Moviesdata/movies
      - /srv/dev-disk-by-uuid-43c6f622-4118-429f-89b7-f9cd3995ea21/RaidSharedBase/Media/Transcodesconfig/transcodes
    ports:
      - 192.168.1.101:8096:8096
      - 192.168.1.101:8920:8920 #optional for secure connections
    restart: unless-stopped
    runtime: nvidia
    deploy:
      resources:
        reservations:
          devices:
            - capabilities: [gpu]

The entire config is located in the pastebin below if you need it. Not all of these services are actually in use, though they're all running. I never got around to setting some of them up.

https://pastebin.com/u5KBa4pQ

[TheDreadPirate]

Update your "Deploy" section to this.

Code:

deploy:
  resources:
    reservations:
      devices:
          - driver: nvidia
             count: all
             capabilities: [gpu]

[SNChalmers]

It works! Thanks so much man, you're a lifesaver. So was it trying to use a different driver or something? I wonder why that would've changed, this stuff worked fine a few months ago.