Omniverse +DoMINO Automotive Aero NIM deployment Web Streaming issue

[preethampai]

Operating System:
Windows
Linux
Kit Version:
107 (Kit App Template)
106 (Kit App Template)
105 (Launcher)
Kit Template:
USD Composer
USD Explorer
USD Viewer
Custom
GPU Hardware:
A series (Blackwell)
A series (ADA)
A series
50 series
40 series
30 series
GPU Driver:
Latest
Recommended (573.xx)
Other
Work Flow:
Omniverse DoMINO Automotive Aero NIM deployment Web Streaming issue

WSL2 with CUDA, NVIDIA-SMI screenshot

Main Issue:
libGLX_nvidia.so.0 || GLX not found in cache, depsite "nvidia-smi " detecting the card.
Reproduction Steps:
Error Code:

[Richard3D]

@preethampai

libGLX_nvidia.so.0 || GLX not found in cache, depsite "nvidia-smi " detecting the card.

This specific issue can occur on Linux systems, especially when:

• The NVIDIA drivers are either not fully installed or incorrectly configured.
• The libGLX_nvidia.so.0 library (essential for OpenGL via NVIDIA drivers) cannot be loaded. This may be due to missing/incorrect symbolic links, or files not present in expected directories.
• Your system is falling back to software rendering (like Mesa/llvmpipe), which won’t work for high-performance graphics or Nvidia-specific workloads like Omniverse streaming.

Common causes and solutions:

• You may need to ensure you used the NVIDIA driver installer correctly, and not a partial install or an outdated version.
• Check that packages like libglx-nvidia0, nvidia-driver, and mesa-utils are properly installed.
• You may need to set the right GLX provider using utilities such as update-glx --config nvidia or similar, depending on your Linux distro.
• Make sure your container (if running via Docker) uses the NVIDIA runtime (--gpus all or --runtime nvidia), as libraries won’t be loaded otherwise.
• Verify paths and symbolic links to libGLX_nvidia.so.0 and related libraries are correct (typically /usr/lib/nvidia/ or /usr/lib/x86_64-linux-gnu/nvidia/).

Bottom line:
Your GPU is detected at a hardware level, but the required NVIDIA OpenGL libraries for streaming/3D acceleration are not found by the software. Fixing your NVIDIA driver installation and GLX setup should resolve it.

[preethampai]

Dear @Richard3D,
Sure, noted.. I did the following check on the following you mentioned: - any more checks ?

@Richard3D

• Check that packages like libglx-nvidia0, nvidia-driver, and mesa-utils are properly installed.
• You may need to set the right GLX provider using utilities such as update-glx --config nvidia or similar, depending on your Linux distro.
• Make sure your container (if running via Docker) uses the NVIDIA runtime (--gpus all or --runtime nvidia), as libraries won’t be loaded otherwise.
• Verify paths and symbolic links to libGLX_nvidia.so.0 and related libraries are correct (typically /usr/lib/nvidia/ or /usr/lib/x86_64-linux-gnu/nvidia/).

also

found a trail from Github: [BUG]: Blank webpage with CORS error and send failure · Issue #8 · NVIDIA-Omniverse-blueprints/digital-twins-for-fluid-simulation

Final push script seems to be : - does this look fine.
docker run --rm --runtime=nvidia --gpus all --shm-size=2g
-p 8000:8000
-e NGC_API_KEY=“$NGC_API_KEY”
nvcr.io/nim/nvidia/domino-automotive-aero:1.0.0

[preethampai]

@Richard3D ,
I’ve been able to get CUDA up running on Ubuntu ..

also installed mesa, libglx, nvidia drivers .. up fresh ..
vulkaninfo --summary

CTK

NCTK

docker run --rm --runtime=nvidia --gpus 1 --shm-size 2g
-p 8000:8000
-e NGC_API_KEY
-t nvcr.io/nim/nvidia/domino-automotive-aero:2.0.0

Docker compose up -d

however Kit app still doesnt recognise the GPU compute

1. is there something that is pushing some defaults in the Kit / app compose.yml ?
2. can we have a unit test to try push only Kit-app using GPU ? without running the whole digital twin ..

any help is highly appreciated …

Thanks in advance ..

[preethampai]

Cuda 13 picked too ..

[preethampai]

@ludokun , how did you get past the web streaming issue highlighted in multiple threads..

any thoughts highly appreciated.

thanks in advance.

[preethampai]

@Richard3D : do you propose anything that may help to debug graphics dependencies

external graphics card detection by WSL2- Ubuntu

[preethampai]

Hello Richard,

I’m unable to login into the developer forum. Can you help me with this? Or who can?

Cheers,
-Preetham Pai N

[Richard3D]

Hi there. You do not need a login, but just to join our Developer Discord channel at Discord

[Richard3D]

I am not clear as to your issues. I think we need to take this back to first principals. Please ignore the Automotive NIM for now.

1. Install the latest NVIDIA Drivers for your Ubuntu OS.
2. Go to our GitHub platform and install our latest kit app template. Run that and install template 2) USD Composer. Get that running first. GitHub - NVIDIA-Omniverse/kit-app-template: Omniverse Kit App Template

Report back with the successful installation and launch and use of USD Composer. Then we can expand the brief.