How to get intel igpu working for local server

[kraizelburg]

Hi, I have vessel installed on my zimaboard2 and it’s working fine but I cannot get the igpu to show any stats.
The server is running Debian headless (no desktop) intel igpu is recognized in /dev/dri as card0 and renderD128

in the beszel agent I have used the “-intel” docker image and passed the device dev/dri as well as the other perfmon parameter as per beszel doc.
The container initiates but get error reading gpu

2026/06/04 16:23:54 WARN Error collecting Intel
GPU data; see https://beszel.dev/guide/gpu err="no valid GPU data found"

intel_gpu_top works as expected on the server but if I run the same inside the container I get

Failed to initialize PMU! (Permission denied)

any ideas? Thanks

[svenvg93]

Can you share your docker compose or service unit for your agent?

[kraizelburg]

Hi, yes sure:

beszel-agent:
    image: henrygd/beszel-agent-intel:latest
    container_name: beszel-agent
    restart: unless-stopped
    network_mode: host
    volumes:
      - ./beszel_agent_data:/var/lib/beszel-agent
      - ./beszel_socket:/beszel_socket
    environment:
      DOCKER_HOST: tcp://ip:2375
      LISTEN: /beszel_socket/beszel.sock
      HUB_URL: http://ip:8095
      TOKEN: tokenid
      KEY: 'sshkey'
      DISABLE_SSH: true
      GPU_COLLECTOR: intel_gpu_top
    cap_add:
     - CAP_PERFMON
    devices:
     - /dev/dri:/dev/dri

Please let me know if you find something odd.

[svenvg93]

Only thing compared to the docs is the devices section that misses the card0 for example. You can us ls /dev/dri to find the name of your GPU.
Might be that you need to set the perf_event_paranoid (docs)on the host system.

[kraizelburg]

Hi, it should pick up the card just passing the /dev/dri and I only have 1 igpu so it’s card0 100% but monitor it’s not plugged it’s a server, only device using the igpu is renderD128. Does beszel doesn’t support monitoring renderD128 with intel_gpu_top? I thought it does, even it’s on the docs. Enviado desde mi iPhone El 09.06.2026, a las 20:18, Sven van Ginkel ***@***.***> escribió:  Only thing compared to the docs is the devices section that misses the card0 for example. You can us ls /dev/dri to find the name of your GPU. Might be that you need to set the perf_event_paranoid (docs<https://beszel.dev/guide/gpu#intel-troubleshooting>)on the host system. — Reply to this email directly, view it on GitHub<#2059?email_source=notifications&email_token=ATL6H2DI6XVZMCV4U7CCOOT47BIG7A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZSGQYDOMRSUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVRTG633UMVZF6Y3MNFRWW#discussioncomment-17240722>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ATL6H2BN3EKHP4J6D3MB65T47BIG7AVCNFSM6AAAAACZ36VMWKVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTOMRUGA3TEMQ>. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS<https://github.com/notifications/mobile/ios/ATL6H2FUMY6QLXRYVRHOMD347BIG7A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZSGQYDOMRSUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVJTG633UMVZF62LPOM> and Android<https://github.com/notifications/mobile/android/ATL6H2DPIRYRB3OJRVKXPZL47BIG7A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZSGQYDOMRSUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVZTG633UMVZF6YLOMRZG62LE>. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>