Nvidia gpu resource Monitor with a docker container

TrueNAS General
1

Hi everyone,

i’ve been waiting for close to 2 years now for a way to monitor my nvidia gpu usage from the gui without using the shell and nvidia-smi. Around those 2 years there were first drafts for a gui widget, but since then no news when we can expect the release, and recent feature requests got closed.

Today at work i was bored and was browsing reddit (don’t judge me) when i stumbled upon a project that displays metrics for nvidia gpus in a webui. It looks like this

image
image952×826 72.3 KB

The only thing i can’t figure out is why the power usage won’t display for my T400 (you may not encounter this).

For anyone interested this is the yaml i used to deploy it as a custom app from the webui:

services:
  gpu-monitor:
    image: bigsk1/gpu-monitor:latest
    ports:
      - "8089:8081" #adjust ports as needed
    volumes:
      - /etc/localtime:/etc/localtime:ro  # Share host timezone
      - /mnt/Plugins/docker/data/gpumon/history:/app/history:rw    # adjust path to a local path on your data pool
      - /mnt/Plugins/docker/data/gpumon/logs:/app/logs:rw    # adjust path to a local path on your data pool
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    restart: unless-stopped
    runtime: nvidia
x-portals:
  - host: 0.0.0.0 #adjust ip as needed
    name: Web UI
    path: /
    port: 8089 #adjust port as needed
    scheme: http
1 Like
2

You usually only need one.

3

You’re 100% right, i actually didn’t see that it both options were in the yaml