-
Notifications
You must be signed in to change notification settings - Fork 211
Description
Required prerequisites
- I have searched the Issue Tracker that this hasn't already been reported. (comment there if it has.)
- I have tried the latest version of nvitop in a new isolated virtual environment.
Motivation
I admin a server with several NVIDIA A100 GPUs, each divided into 4 MIG GPUS. It is difficult for me to identify which MIG GPU matches with which incus container because I cannot see the MIG UUIDS in the main nvitop monitor display.
Solution
I would like there to be a option so that the MIG UUIDs could be displayed in full next to the GPU numeric ID column in the main nvitop monitor terminal/ssh display.
Alternatives
The nvitop docs say that I should be able to navigate between processes in the (full) monitor display, running via ssh but I have been unable to select any process / GPU to inspect the details of a process. I have been pushing the UP and DOWN arrow keys, J, K, Alt-J, Alt-K, HOME, END... nothing. No processes get highlighted. I am running it as the root user.
The server is running the latest nvitop as ran via pipx under proxmox Debian 13.
The client is Windows 10 running Windows ssh in the new Windows Terminal ie not W10 command prompt.
So because of this, I haven't been able to view any MIG UUIDS from within nvitop.
I have tried nvidia-smi but nvitop seems like it will be better suited and its open source.
nvidia-smi provides almost no monitoring features when run inside incus containers with a MIG GPU attached but nvitop does work in this environment.
Additional context
No response