utkuozdemir / nvidia_gpu_exporter

Nvidia GPU exporter for prometheus using nvidia-smi binary

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Can't visualize metrics on Grafana Dashboard

mbenencase opened this issue · comments

Hey, currently I'm facing a problem that I can't visualize the metrics on my Grafana Dashboard. I'll describe the steps that I followed

  1. I installed the .deb package according to what is described in INSTALL.md. My laptop has Ubuntu 20.04 and my GPU is a GeForce 3060.
  2. On CONFIGURE.md I set the command of nvidia-smi as
nvidia_gpu_exporter --nvidia-smi-command 'nvidia-smi'
  1. I imported the dashboard that is ready-to-use on Grafana, but no metrics appeared.

After step (3), I was looking in this GitHub for similar issues and I found this one: #7 . However, one of the suggestions is to verify the metrics log on Prometheus on http://localhost:9090, but this page throws me the 404 Error Page not Found. So I believe that there's a Prometheus setup step that I'm missing. And that's what I think it's my problem.

How could I setup Prometheus properly to be able to visualize the metrics on the Dashboard?

Hi,

You do not need to pass --nvidia-smi-command nvidia-smi since it is the default value anyway - the flag is only for the cases when you want to override it.

The problem indeed seems like related to your Prometheus setup. Are you running on Windows?

Hey, thanks for the first comment, I didn't know about that.

About the question, no, currently I'm on Ubuntu 20.04

Cool, how did you install prometheus? Is it installed as a systemd service?

You can check its logs using journalctl -xeu prometheus (replace with the service name it is installed, if it is not prometheus).

Your first goal must be to get http://localhost:9090 working and seeing some metrics there. You need to troubleshoot your Prometheus installation. Only then this exporter becomes relevant.

@utkuozdemir I think that my problem is that I didn't install it. I'll verify how to install it properly.

Hm, then it's weird that you got 404 Not Found and not Connection Refused. Some process must be listening port 9090 on your system already.

Closing this issue, as it is not related to the exporter itself.