GPU plugin do (can) not work on Alpine image

[yeeahnick]

Hello,

I'm encountering an issue where my NVIDIA Quadro P4000 is not being detected by Glances. I'm using the docker-compose (latest-full) configuration and have enabled NVIDIA GPU support in the application settings while building the app in TrueNAS. This configuration sets the NVIDIA_VISIBLE_DEVICES and NVIDIA_DRIVER_CAPABILITIES variables.

With these settings, I can see the NVIDIA driver listed under the file system pane in Glances, but the GPU does not appear when I access the endpoint:
http://IP:61208/api/4/gpu.

Interestingly, when I navigate to http://IP:61208/api/4/full, I can see several NVIDIA-related entries.

To ensure the GPU is properly assigned in the Docker Compose configuration, I ran the following command in the TrueNAS shell:

midclt call -job app.update glances-custom '{"values": {"resources": {"gpus": {"use_all_gpus": false, "nvidia_gpu_selection": {"PCI_SLOT": {"use_gpu": true, "uuid": "GPU-95943d54-8d67-b91e-00cb-ca3662cfd863"}}}}}}'

Despite this, the GPU still doesn’t show up in the /gpu endpoint.

Does anyone have suggestions or insights on what might be missing or misconfigured? Any help would be greatly appreciated!

Thank you!

[nicolargo]

Hi @yeeahnick

can you copy/paste the result of a curl on http://ip:61208/api/4/full ?

Thanks.

[yeeahnick]

Hi @nicolargo

Thanks for the quick response.

Unfortunately the curl /full no longer shows the NVIDIA gpu (same thing under file system in Glances). There was a TrueNAS Scale update (24.10.2) yesterday that included NVIDIA fixes which I guess made it worst for Glances. To be clear my GPU is working in other dockers running on the same system.

But I can give more information.

When I run "ls /dev | grep nvidia" in the shell of Glances I see the following:

nvidia-caps
nvidia-modeset
nvidia-uvm
nvidia-uvm-tools
nvidia0
nviaiactl

When I do a nvidia-smi nothing is found (this works in other dockers on the same system).

When I run "env" in the shell of Glances I see that the NVIDIA capabilities and devices are enabled. (environment variables)

When I run "glances | grep -i runtime" in the shell of Glances it just hangs.

I will fiddle with it again tonight to see if I can repopulate the curl /full.

Let me know if I need to provide anything else.

Cheers!

[nicolargo]

In the shell of Glances, can you run the following command:

glances -V

It will display the path to the glances.log file.

then run:

glances -d --stdout gpu --stop-after 3

And copy paste:

• the glances.log file (relevants lines)
• output of the command

Thanks !

[XSvirusSAFE]

Having the same issue here. Hope the info within the screenshot can help.

[yeeahnick]

@nicolargo

[yeeahnick]

Having the same issue here. Hope the info within the screenshot can help.

You can run this "cat /tmp/glances-root.log" in the Glances shell to view the log file.

[kbirger]

Same exact results here. I have also noticed that inside the container nvidia-smi reports "not found". Bizarre, because it's there

/app # which nvidia-smi
/usr/bin/nvidia-smi
/app # ls -l /usr/bin/nvidia-smi
-rwxr-xr-x    1 root     root       1068640 Jan 30 04:54 /usr/bin/nvidia-smi
/app # stat /usr/bin/nvidia-smi
  File: /usr/bin/nvidia-smi
  Size: 1068640         Blocks: 2088       IO Block: 4096   regular file
Device: fc06h/64518d    Inode: 9710205     Links: 1
Access: (0755/-rwxr-xr-x)  Uid: (    0/    root)   Gid: (    0/    root)
Access: 2025-01-30 04:55:14.164896111 +0000
Modify: 2025-01-30 04:54:48.715960042 +0000
Change: 2025-01-30 04:54:48.715960042 +0000
/app # nvidia-smi
sh: nvidia-smi: not found
/app # /usr/bin/nvidia-smi
sh: /usr/bin/nvidia-smi: not found

/app # id
uid=0(root) gid=0(root) groups=0(root),1(bin),2(daemon),3(sys),4(adm),6(disk),10(wheel),11(floppy),20(dialout),26(tape),27(video)

from the root log:

2025-01-31 01:57:40,193 -- DEBUG -- NVML Shared Library (libnvidia-ml.so.1) not Found, Nvidia GPU plugin is disabled

However, I've got other containers on the system that are using the GPU no problem.

Please let me know if you want to see any other parts of the log

[yeeahnick]

Same with nvidia-smi

/

[nicolargo]

Glances binds directly the libnvidia-ml.so.1 file. Check that this file is available on your system.

find /usr -name 'libnvidia-ml.so*'

The folder where this file is located should be added to LD_LIBRARY_PATH.

So long story short it's more a TrueNAS integration issue than a Glances bug.

[XSvirusSAFE]

Please see the attached.

[yeeahnick]

Glances binds directly the libnvidia-ml.so.1 file. Check that this file is available on your system.

find /usr -name 'libnvidia-ml.so*'

The folder where this file is located should be added to LD_LIBRARY_PATH.

So long story short it's more a TrueNAS integration issue than a Glances bug.

Bonjour et merci de vous impliquer avec ce problème.

Is there something that can be done with the glances container to fix this? I doubt TrueNAS will take a look at this since all my other dockers an community apps have a working GPU without doing anything special. (immich, plex, mkvtoolnix, dashdot, etc..)

I set LD_LIBRARY_PATH to /usr/lib/x86_64-linux-gnu (also tried with /usr/lib64 ) as env variable on the container but it didn't change anything. Also did the same for LD_PRELOAD.

I also tried the alpine-dev tags but got the same results (I do see more info like IP) . I also tried with the official truenas community app but that one doesn't support GPUs at all.

Here is the result of that command in the Glances shell:

/app # find /usr -name 'libnvidia-ml.so*'
/usr/lib64/libnvidia-ml.so.1
/usr/lib64/libnvidia-ml.so.550.127.05
/app # 

/app # ls /usr/lib64
libEGL_nvidia.so.0                       libcudadebugger.so.550.127.05            libnvidia-glsi.so.550.127.05             libnvidia-opticalflow.so.550.127.05
libEGL_nvidia.so.550.127.05              libnvcuvid.so.1                          libnvidia-glvkspirv.so.550.127.05        libnvidia-pkcs11-openssl3.so.550.127.05
libGLESv1_CM_nvidia.so.1                 libnvcuvid.so.550.127.05                 libnvidia-gpucomp.so.550.127.05          libnvidia-pkcs11.so.550.127.05
libGLESv1_CM_nvidia.so.550.127.05        libnvidia-allocator.so.1                 libnvidia-ml.so.1                        libnvidia-ptxjitcompiler.so.1
libGLESv2_nvidia.so.2                    libnvidia-allocator.so.550.127.05        libnvidia-ml.so.550.127.05               libnvidia-ptxjitcompiler.so.550.127.05
libGLESv2_nvidia.so.550.127.05           libnvidia-cfg.so.1                       libnvidia-ngx.so.1                       libnvidia-rtcore.so.550.127.05
libGLX_indirect.so.0                     libnvidia-cfg.so.550.127.05              libnvidia-ngx.so.550.127.05              libnvidia-tls.so.550.127.05
libGLX_nvidia.so.0                       libnvidia-eglcore.so.550.127.05          libnvidia-nvvm.so.4                      libnvoptix.so.1
libGLX_nvidia.so.550.127.05              libnvidia-encode.so.1                    libnvidia-nvvm.so.550.127.05             libnvoptix.so.550.127.05
libcuda.so                               libnvidia-encode.so.550.127.05           libnvidia-opencl.so.1                    libvdpau_nvidia.so.1
libcuda.so.1                             libnvidia-fbc.so.1                       libnvidia-opencl.so.550.127.05           libvdpau_nvidia.so.550.127.05
libcuda.so.550.127.05                    libnvidia-fbc.so.550.127.05              libnvidia-opticalflow.so                 xorg
libcudadebugger.so.1                     libnvidia-glcore.so.550.127.05           libnvidia-opticalflow.so.1

Here is the result of that command in the TrueNAS shell:

root@truenas[~]# find /usr -name 'libnvidia-ml.so*'
/usr/lib/x86_64-linux-gnu/libnvidia-ml.so
/usr/lib/x86_64-linux-gnu/libnvidia-ml.so.1
/usr/lib/x86_64-linux-gnu/libnvidia-ml.so.550.127.05
root@truenas[~]# 

root@truenas[~]# nvidia-container-cli list
/dev/nvidiactl
/dev/nvidia-uvm
/dev/nvidia-uvm-tools
/dev/nvidia-modeset
/dev/nvidia0
/usr/bin/nvidia-smi
/usr/bin/nvidia-debugdump
/usr/bin/nvidia-persistenced
/usr/bin/nvidia-cuda-mps-control
/usr/bin/nvidia-cuda-mps-server
/usr/lib/x86_64-linux-gnu/libnvidia-ml.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-cfg.so.550.127.05
/usr/lib/x86_64-linux-gnu/libcuda.so.550.127.05
/usr/lib/x86_64-linux-gnu/libcudadebugger.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-opencl.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-gpucomp.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-ptxjitcompiler.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-allocator.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-pkcs11.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-pkcs11-openssl3.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-nvvm.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-ngx.so.550.127.05
/usr/lib/x86_64-linux-gnu/vdpau/libvdpau_nvidia.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-encode.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-opticalflow.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvcuvid.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-eglcore.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-glcore.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-tls.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-glsi.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-fbc.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-rtcore.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvoptix.so.550.127.05
/usr/lib/x86_64-linux-gnu/libGLX_nvidia.so.550.127.05
/usr/lib/x86_64-linux-gnu/libEGL_nvidia.so.550.127.05
/usr/lib/x86_64-linux-gnu/libGLESv2_nvidia.so.550.127.05
/usr/lib/x86_64-linux-gnu/libGLESv1_CM_nvidia.so.550.127.05
/usr/lib/x86_64-linux-gnu/libnvidia-glvkspirv.so.550.127.05
/lib/firmware/nvidia/550.127.05/gsp_ga10x.bin
/lib/firmware/nvidia/550.127.05/gsp_tu10x.bin

root@truenas[~]# nvidia-container-cli --version
cli-version: 1.17.4
lib-version: 1.17.4
build date: 2025-01-23T10:53+00:00
build revision: f23e5e55ea27b3680aef363436d4bcf7659e0bfc
build compiler: x86_64-linux-gnu-gcc-7 7.5.0
build platform: x86_64
build flags: -D_GNU_SOURCE -D_FORTIFY_SOURCE=2 -DNDEBUG -std=gnu11 -O2 -g -fdata-sections -ffunction-sections -fplan9-extensions -fstack-protector -fno-strict-aliasing -fvisibility=hidden -Wall -Wextra -Wcast-align -Wpointer-arith -Wmissing-prototypes -Wnonnull -Wwrite-strings -Wlogical-op -Wformat=2 -Wmissing-format-attribute -Winit-self -Wshadow -Wstrict-prototypes -Wunreachable-code -Wconversion -Wsign-conversion -Wno-unknown-warning-option -Wno-format-extra-args -Wno-gnu-alignof-expression -Wl,-zrelro -Wl,-znow -Wl,-zdefs -Wl,--gc-sections```

[kbirger]

Glances binds directly the libnvidia-ml.so.1 file. Check that this file is available on your system.

find /usr -name 'libnvidia-ml.so*'

The folder where this file is located should be added to LD_LIBRARY_PATH.

So long story short it's more a TrueNAS integration issue than a Glances bug.

I'm not on TrueNAS actually. I'm on Proxmox, which is a fork of debian.

libnvidia-ml.so is found on both my host, and in the container.

From the container: /usr/lib64

total 219456
drwxr-xr-x    2 root     root          4096 Feb  3 15:28 .
drwxr-xr-x    1 root     root          4096 Feb  3 15:28 ..
lrwxrwxrwx    1 root     root            12 Feb  3 15:28 libcuda.so -> libcuda.so.1
lrwxrwxrwx    1 root     root            21 Feb  3 15:28 libcuda.so.1 -> libcuda.so.550.144.03
-rwxr-xr-x    1 root     root      28712096 Jan 30 04:54 libcuda.so.550.144.03
lrwxrwxrwx    1 root     root            29 Feb  3 15:28 libcudadebugger.so.1 -> libcudadebugger.so.550.144.03
-rwxr-xr-x    1 root     root      10524136 Jan 30 04:54 libcudadebugger.so.550.144.03
lrwxrwxrwx    1 root     root            33 Feb  3 15:28 libnvidia-allocator.so.1 -> libnvidia-allocator.so.550.144.03
-rwxr-xr-x    1 root     root        168808 Jan 30 04:54 libnvidia-allocator.so.550.144.03
lrwxrwxrwx    1 root     root            27 Feb  3 15:28 libnvidia-cfg.so.1 -> libnvidia-cfg.so.550.144.03
-rwxr-xr-x    1 root     root        398968 Jan 30 04:54 libnvidia-cfg.so.550.144.03
-rwxr-xr-x    1 root     root      43659040 Jan 30 04:54 libnvidia-gpucomp.so.550.144.03
lrwxrwxrwx    1 root     root            26 Feb  3 15:28 libnvidia-ml.so.1 -> libnvidia-ml.so.550.144.03
-rwxr-xr-x    1 root     root       2082456 Jan 30 04:54 libnvidia-ml.so.550.144.03
lrwxrwxrwx    1 root     root            28 Feb  3 15:28 libnvidia-nvvm.so.4 -> libnvidia-nvvm.so.550.144.03
-rwxr-xr-x    1 root     root      86842616 Jan 30 04:54 libnvidia-nvvm.so.550.144.03
lrwxrwxrwx    1 root     root            30 Feb  3 15:28 libnvidia-opencl.so.1 -> libnvidia-opencl.so.550.144.03
-rwxr-xr-x    1 root     root      23613128 Jan 30 04:54 libnvidia-opencl.so.550.144.03
-rwxr-xr-x    1 root     root         10176 Jan 30 04:54 libnvidia-pkcs11-openssl3.so.550.144.03
-rwxr-xr-x    1 root     root         10168 Jan 30 04:54 libnvidia-pkcs11.so.550.144.03
lrwxrwxrwx    1 root     root            38 Feb  3 15:28 libnvidia-ptxjitcompiler.so.1 -> libnvidia-ptxjitcompiler.so.550.144.03
-rwxr-xr-x    1 root     root      28674464 Jan 30 04:54 libnvidia-ptxjitcompiler.so.550.144.03
/app #3003 

I assume that you meant taht the env var must be set to the path in the container, otherwise it would also be necessary to mount the file from the host.

Setting it to /usr/lib64 doesn't make a difference

[yeeahnick]

@nicolargo

Hi,

Can we change the label "needs more info" to "need investigation".

A few of us provided a lot of info and we all have the same results and issue. TrueNAS and Proxmox are affected.

Thanks

[kbirger]

@nicolargo it looks like you are swallowing the original exception (https://github.com/nicolargo/glances/blob/develop/glances/plugins/gpu/cards/nvidia.py#L32), but if I shell into the container and run the code out of nvidia.py, I get this error:

OSError: Error relocating /usr/lib64//libnvidia-ml.so.1: dlvsym: symbol not found

the double slash is strange to me. The file exists here

lrwxrwxrwx    1 root     root            26 Feb 13 03:18 /usr/lib64/libnvidia-ml.so.1 -> libnvidia-ml.so.550.144.03

[nicolargo]

@kbirger

Id do not have any computer with a NVIDIA card, so i need you help to investigate.

Can you open a Python shell and enter the following command:

> python
...
>>> from ctypes import CDLL
>>> CDLL('libnvidia-ml.so.1')
...
>>> CDLL('libnvidia-ml.so.550.144.03')
...

Please copy/paste the result.

Thanks.

[yeeahnick]

@nicolargo I can confirm what @kbirger said is accurate. The ubuntu-latest-full detects my P4000 no problem. Looks like the issue is within the latest-full (Alpine).

[dbsavage]

@nicolargo I actually tried CDLL on the first file. I think I got the same error as above. I'll try again on Monday or Tuesday, as I won't have access to a machine until then.

I have also been meaning to mention that this is only an issue on the alpine image. I switched to the other and have no problems. I'd still like to help you track down the issue with this one, because I'd prefer to run alpine for the smaller footprint. I mention it only in case it helps track down the issue.

I also switched to the Ubuntu based container instead of the alpine one and things started to work for me. Note: I'm on Ubuntu server 24.04 and not proxmox or truenas. But the alpine image does not work for monitoring gpu for me... Nvidia gpu. Ubuntu based container works fine.

[nicolargo]

I can not reproduce the issue on my side, so i need you to investigate.

First of all, identify the current NVidia lib with the current command:

find /usr -name libnvidia-ml.so.1

It should return one file with a full path.

With the output of the previous command (do not forget the * at the end):

ls -alF <output ot the first command>*

It should return a minimum of 2 files (on is a symbolic link and another one is the target file).

For each line from the second command:

python
> from ctypes import CDLL
> CDLL('<output of the second command>')

Please copy/paste all the results.

[yeeahnick]

Like this?

>>> /app # find /usr -name libnvidia-ml.so.1
/usr/lib64/libnvidia-ml.so.1
/app # ls -alF /usr/lib64/libnvidia-ml.so.1
lrwxrwxrwx    1 root     root            26 Mar  2 12:17 /usr/lib64/libnvidia-ml.so.1 -> libnvidia-ml.so.550.127.05*
/app # python
Python 3.12.8 (main, Dec  7 2024, 05:56:13) [GCC 14.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from ctypes import CDLL
>>> CDLL ('/usr/lib64/libnvidia-ml.so.1 -> libnvidia-ml.so.550.127.05*')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.12/ctypes/__init__.py", line 379, in __init__
    self._handle = _dlopen(self._name, mode)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^
OSError: Error loading shared library /usr/lib64/libnvidia-ml.so.1 -> libnvidia-ml.so.550.127.05*: No such file or directory
>>> 

[nicolargo]

Nope.

You need to add the * to also see the file targeted by the symbolic link

ls -alF /usr/lib64/libnvidia-ml.so.1*

And apply the CDLL command on each file:

Some thing like that:

CDLL ('/usr/lib64/libnvidia-ml.so.1')
CDLL (/'usr/lib64/libnvidia-ml.so.550.127.05')

Thanks !

[yeeahnick]

Sorry if I am doing it wrong again. Here is what I got this time.

/app # find /usr -name libnvidia-ml.so.1
/usr/lib64/libnvidia-ml.so.1
/app # ls -alF /usr/lib64/libnvidia-ml.so.1*
lrwxrwxrwx    1 root     root            26 Mar  2 13:08 /usr/lib64/libnvidia-ml.so.1 -> libnvidia-ml.so.550.127.05*
/app # python
Python 3.12.8 (main, Dec  7 2024, 05:56:13) [GCC 14.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from ctypes import CDLL
>>> CDLL ('/usr/lib64/libnvidia-ml.so.1')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.12/ctypes/__init__.py", line 379, in __init__
    self._handle = _dlopen(self._name, mode)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^
OSError: Error relocating /usr/lib64/libnvidia-ml.so.1: dlvsym: symbol not found
>>> CDLL ('/usr/lib64/libnvidia-ml.so.550.127.05')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.12/ctypes/__init__.py", line 379, in __init__
    self._handle = _dlopen(self._name, mode)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^
OSError: Error relocating /usr/lib64/libnvidia-ml.so.550.127.05: dlvsym: symbol not found
>>> 

[nicolargo]

It's strange.

The following line

app # ls -alF /usr/lib64/libnvidia-ml.so.1*

should return 2 files:

/usr/lib64/libnvidia-ml.so.1 -> libnvidia-ml.so.550.127.05*

and

/usr/lib64/libnvidia-ml.so.550.127.05

Can you also copy paste:

app # ls -alF  /usr/lib64/libnvidia-ml.so.550.127.05

[yeeahnick]

Here you go and thanks for looking into this.

FYI the ubuntu-latest-full that is working also only shows 1 file.

/app # ls -alF  /usr/lib64/libnvidia-ml.so.550.127.05
-rwxr-xr-x    1 root     root       2078360 Jan 27 23:19 /usr/lib64/libnvidia-ml.so.550.127.05*
/app # 

[nicolargo]

So the file exist but can not be loaded as a proper lib...

What's the libnvidia-ml version on your working Ubuntu image ? (the same than the Alpine: 550.127.05) ?

[yeeahnick]

Yes same version but location is '/usr/lib/x86_64-linux-gnu'

[alexfornuto]

Sorry if this is noise in the thread, but I'll add that I was having the same issues on a Debian server, using Glances in Docker. Glances only saw the AMD graphics integrated in the CPU. I switched from the latest tag to ubuntu-4.3.0.8-full, and now I can see my nvidia card and stats.

[cyberclaw03]

Glances only saw the AMD graphics integrated in the GPU. I switched from the latest tag to ubuntu-4.3.0.8-full, and now I can see my nvidia card and stats.

I can confirm that this worked for me as well on Pop OS. Nothing else enabled me to see the NVIDIA GPUs

[marktuk]

I was also having this problem on TrueNAS Scale 24.10 with a Quadro P620 installed, but after switching my docker-compose file to use the ubuntu-latest-full image I can see my GPU in glances.

[Sterahi]

Adding my config to the mix.

I'm running glances through Docker on an Ubuntu 22.04.4 machine.

Host Machine: Ubuntu 22.04.4
Configuration: Docker (through Portainer)
GPU: Nvidia 1660ti (driver version: 535.183.01)

Working docker-compose file:

version: '3'

services:
  monitoring:
    image: nicolargo/glances:ubuntu-latest-full
    pid: host
    network_mode: host
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock
      # Uncomment the below line if you want glances to display host OS detail instead of container's
      # - /etc/os-release:/etc/os-release:ro
    environment:
      - "GLANCES_OPT=-w"

Not working (lifted directly from the docs):

version: '3'

services:
  monitoring:
    image: nicolargo/glances:latest-full
    pid: host
    network_mode: host
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock
      # Uncomment the below line if you want glances to display host OS detail instead of container's
      # - /etc/os-release:/etc/os-release:ro
    environment:
      - "GLANCES_OPT=-w"
    # For nvidia GPUs
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

I noticed some interesting things while getting my GPU to finally work, some of them might be docker/portainer specific issues or just something weird I did on my server.

I noticed that including the deploy block prevented me from assigning my GPU to the container so I removed it in favor of assigning it in Portainer. This issue happened with both the Ubuntu & Alpine full images, it's almost certainly something on my system.

Some notes that might help debug this further:

• alpine does not have a lib64 directory.
  • ubuntu has a lib64 directory with a symlink to ld-linux-x86-64.so.2 -> ../lib/x86_64-linux-gnu/ld-linux-x86-64.so.2
  • I checked the lib directory on both alpine & ubuntu, alpine is completely missing the libnvidia files that ubuntu has.
  • Previously provided python code confirms this, ubuntu can find the file, alpine cannot.

[yeeahnick]

Chatgpt tells me that Alpine does not support Nvidia drivers out of the box. With that in mind the missing Nvidia Lib is probably normal. Hence why the GPU only works with the Ubuntu image.

[alexfornuto]

Seems like the thing to do then is add an install step for the lib to the build file?