[Bug]: nvtop + intel i915 gpus doc missing fields

[incaseoftrouble]

Welcome!

• I have read the Documentation
• I have checked the Common Issues Guide and my problem was not mentioned there.
• I have searched open and closed issues and my problem was not mentioned before.
• I have verified I am using the latest version available. You can check the latest release here.

Component

Agent

Problem Description

I have two Intel GPUs (i915) -- 12400 iGPU + Arc 380. I want to use nvtop to get dual-gpu support and memory reporting. I'm using docker with henrygd/beszel-agent-intel

I have pid: host and GPU_COLLECTOR: nvtop in the docker compose file, as per doc.

Two issues arise:

• nvtop -s in the container does not report the names ("device_name": null) unless I add /run/udev:/run/udev:ro as mount (or at least /run/udev/data) (the nvtop process also seems to try to access /etc/udev/udev.conf but not having it seems harmless)
• nvtop -s does not report the utilization and processes on the second GPU unless I add SYS_PTRACE (the first card's processes do appear though!)

I suggest adding this to the doc.

Average utilization of GPU engines seems to be empty though (it did report on load on my A380 using intel_gpu_top). nvtop does report encode_decode. Maybe this is an artifact of using intel_gpu_top before? nvtop doesn't have the same granularity I suppose -- yet, reporting encode_decode would be cool, its the key metric for transcoding.

EDIT: Reading dmesg I now understand at least the PTRACE part is due to apparmor. Adding

security_opt:
  - apparmor:unconfined
  - seccomp:unconfined

seems to also be required

Expected Behavior

GPUs appear in hub

Steps to Reproduce

Run agent in docker on a dual intel gpu machine.

OS / Architecture

linux/amd64 (agent) on docker, debian trixie 6.12.101 kernel

Beszel version

0.18.8

Installation method

Docker

Configuration

Hub Logs

Agent Logs

[svenvg93]

@henrygd can you move this one to the docs repo :) (I don't have the access to do that)