0

I am trying to get load statistics from my GPUs (NVIDIA). Is there a place in the /proc file system where I could find something useful, akin to /proc/stat to get CPU info?

I know there are command line tools to do this, but in my particular case these will not do as I will run the monitor in a container. I can mount /proc from the host to my monitoring container and access anything that is in there, but I cannot run commands on the host.

How would I do this, or is this something that cannot be done?

Hannu
  • 113
  • 5
  • tried 'nvidia-smi'? – Chopper3 Jun 15 '18 at 14:28
  • Does not work from a docker container. This is why I need to find something in /proc I can read. – Hannu Jun 15 '18 at 14:33
  • Of course could be another file system location as well but thought /proc might be the most promising place to look. – Hannu Jun 15 '18 at 14:44
  • I didn't even know you could talk directly to a GPU from Docker, you learn something new every day! – Chopper3 Jun 15 '18 at 16:01
  • I do not know if you can, nor do I care. I am only running my monitoring script in a container for reasons not relevant to this. The script is about 100 lines of very standard Python. My question is, can I get this info from the /proc file system or not. – Hannu Jun 17 '18 at 17:43

0 Answers0