nvidia-smi: command not found, but GPU works fine

Question

I was trying to install some package in my server and that requires some updates on some CUDA libraries. But now I end up getting

nvidia-smi: command not found,

even my GPUs works fine like before. So what could potentially mess up and how can I solve it? Some info:

OS: Ubuntu 16.04

$ lspci | grep VGA
04:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)
05:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)
08:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)
09:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)
0c:00.0 VGA compatible controller: ASPEED Technology, Inc. ASPEED Graphics Family (rev 30)
84:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)
85:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)
88:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)
89:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)


$ dpkg -l | grep -E "nvidia-[0-9]{3}"
rc  nvidia-390                         390.59-0ubuntu0~gpu16. amd64                  NVIDIA binary driver - version 390.59
rc  nvidia-396                         396.54-0ubuntu0~gpu16. amd64                  NVIDIA binary driver - version 396.54
iU  nvidia-440                         440.33.01-0ubuntu1     amd64                  NVIDIA binary driver - version 440.33.01
iU  nvidia-440-dev                     440.33.01-0ubuntu1     amd64                  NVIDIA binary Xorg driver development files

$ grep "X Driver" /var/log/Xorg.0.log
[  4612.223] (II) NVIDIA dlloader X Driver  396.54  Tue Aug 14 22:15:03 PDT 2018

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Nov__3_21:07:56_CDT_2017
Cuda compilation tools, release 9.1, V9.1.85

Question is can I solve the nvidia-smi: command not found problem (which didn't happen before I was trying install some package) without apt purge nvidia-* or install any other ppa? It's a shared server so I don't want to mess up anything

Solution

nvidia-smi is installed via nvidia-utils, as shown here:

$ sudo apt-get install nvidia-smi     
Reading package lists... Done
Building dependency tree       
Reading state information... Done
Package nvidia-smi is a virtual package provided by:
  nvidia-utils-440 440.44-0ubuntu0~0.18.04.1
  nvidia-utils-435 435.21-0ubuntu0.18.04.2
  nvidia-utils-430 430.64-0ubuntu0~gpu18.04.1
  nvidia-utils-415 415.27-0ubuntu0~gpu18.04.2
  nvidia-utils-410 410.104-0ubuntu0~18.04.1
  nvidia-utils-390 390.129-0ubuntu0~gpu18.04.1
You should explicitly select one to install.

E: Package 'nvidia-smi' has no installation candidate

So, in your case you'll need to install nvidia-utils-440.

Comments

[Sadaharu Wakisaka]

Welcome to AskUbuntu, please describe your question clearly. I don't understand what you want to ask here.

[Zepeng Huo]

@SadaharuWakisaka thanks for your comment. I do realize it was vague. So I'm asking what could mess up nvidia-smi utility and how can I solve that problem?

[singrium]

Is secure boot disabled from the BIOS menu? If not, disable it.

[nobody]

first of all fix your packagemanagement . iU nvidia-440 iU nvidia-440-dev At least these 2 packages are not full installt.

[Zepeng Huo]

Aren't all of them showing ubuntu 18.04? But mine is 16.04?

[ajgringo619]

Should have clarified - this list is from my system (Mint 19.2), so it's showing 18.04 packages. Just run sudo apt install nvidia-utils-440 and you'll get the right version.