Can the Official Plex Docker use NVIDIA GPU Transcoding through Compose?

[yottabit]

I have searched and searched, and I cannot figure out why my official Plex Docker compose is not using the GPU for transcoding. I’m using Dockge to manage. Here is the compose file:

services:
  pms-docker:
    container_name: plex
    ports:
      - 32400:32400/tcp
      - 3005:3005/tcp
      - 8324:8324/tcp
      - 32469:32469/tcp
      - 1900:1900/udp
      - 32410:32410/udp
      - 32412:32412/udp
      - 32413:32413/udp
      - 32414:32414/udp
    environment:
      - ADVERTISE_IP=http://172.16.42.139:32400/
      - ALLOWED_NETWORKS=192.168.0.0/16,172.16.0.0/12
      - NVIDIA_VISIBLE_DEVICES=${GPU_DEVICE_ID}
      - PLEX_CLAIM=claim...
      - PLEX_UID=568
      - PLEX_GID=568
      - TZ=America/Chicago
    hostname: rwithrow-plex
    volumes:
      - ${DOCKER_ROOT}/config/plex:/config
      - /dev/shm:/transcode
      - ${DOCKER_ROOT}/data/plex:/data
      - ${POOL_ROOT}/tv:/mnt/tv
      - ${POOL_ROOT}/movies:/mnt/movies
      - ${POOL_ROOT}/audiobooks:/mnt/audiobooks
    image: plexinc/pms-docker:plexpass
    runtime: nvidia
    restart: unless-stopped
networks: {}

And .env:

DOCKER_ROOT=/mnt/pool1/docker
GPU_DEVICE_ID=0000:02:00.0
POOL_ROOT=/mnt/pool1

The TrueNAS repo Plex was using the GPU successfully. Things I have tried:

1. Many variations in the compose file
2. Disabling and then enabling the GPU setting in TrueNAS Apps

Running nvidia-smi in the container shows my Quadro P400. And Plex has “Use hardware-accelerated video encoding” enabled.

What am I missing? Does the official Plex Docker just not work with Docker compose? That would seem odd. I wish we could see the compose file used by TrueNAS apps. I have seen an example on this forum by Tyler_Shield, but he is using a third-party Docker image.

[LarsR]

for my jellyfin im using this in my compose

    runtime: nvidia
    deploy:
         resources:
            reservations:
              devices:
                - driver: nvidia
                  count: 1
                  capabilities: [gpu] 

instead of the env variables

Edit: you do have a Plex pass right? Just to make sure

[yottabit]

Thanks for the idea! Unfortunately it didn’t make a difference for me. And yes, I do have a plexpass.

Updated compose file:

services:
  pms-docker:
    container_name: plex
    ports:
      - 32400:32400/tcp
      - 3005:3005/tcp
      - 8324:8324/tcp
      - 32469:32469/tcp
      - 1900:1900/udp
      - 32410:32410/udp
      - 32412:32412/udp
      - 32413:32413/udp
      - 32414:32414/udp
    environment:
      - ADVERTISE_IP=http://172.16.42.139:32400/
      - ALLOWED_NETWORKS=192.168.0.0/16,172.16.0.0/12
      - NVIDIA_VISIBLE_DEVICES=${GPU_DEVICE_ID}
      - PLEX_CLAIM=claim...
      - PLEX_UID=568
      - PLEX_GID=568
      - TZ=America/Chicago
    hostname: rwithrow-plex
    volumes:
      - ${DOCKER_ROOT}/config/plex:/config
      - /dev/shm:/transcode
      - ${DOCKER_ROOT}/data/plex:/data
      - ${POOL_ROOT}/tv:/mnt/tv
      - ${POOL_ROOT}/movies:/mnt/movies
      - ${POOL_ROOT}/audiobooks:/mnt/audiobooks
    image: plexinc/pms-docker:plexpass
    runtime: nvidia
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu] 
    restart: unless-stopped
networks: {}

[yottabit]

Just found this on a blog post:

The official image is missing NVIDIA packages that prevent transcoding from working correctly.

I guess TrueNAS is not using the official image, either!

[yottabit]

Yessss, it is now working with a different image!!

This is the working compose file:

services:
  pms-docker:
    container_name: plex
    ports:
      - 32400:32400/tcp
      - 3005:3005/tcp
      - 8324:8324/tcp
      - 32469:32469/tcp
      - 1900:1900/udp
      - 32410:32410/udp
      - 32412:32412/udp
      - 32413:32413/udp
      - 32414:32414/udp
    environment:
      - ADVERTISE_IP=http://172.16.42.139:32400/
      - ALLOWED_NETWORKS=192.168.0.0/16,172.16.0.0/12
      - NVIDIA_DRIVER_CAPABILITIES=compute,video,utility
      - NVIDIA_VISIBLE_DEVICES=all
      - PLEX_CLAIM=claim-...
      - PUID=568
      - PGID=568
      - TZ=America/Chicago
      - VERSION=docker
    hostname: rwithrow-plex
    volumes:
      - ${DOCKER_ROOT}/config/plex:/config
      - /dev/shm:/transcode
      - ${DOCKER_ROOT}/data/plex:/data
      - ${POOL_ROOT}/tv:/mnt/tv
      - ${POOL_ROOT}/movies:/mnt/movies
      - ${POOL_ROOT}/audiobooks:/mnt/audiobooks
    image: lscr.io/linuxserver/plex:latest
    runtime: nvidia
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    restart: unless-stopped
networks: {}