[Support Request] NemoClaw Installation Error on Jetson AGX Orin (JetPack 7.2.1)

Robotics & Edge Computing Jetson Systems Jetson AGX Orin
, ,
1

Hello NVIDIA/NemoClaw Support Team,

I am encountering an issue while trying to install NemoClaw on my NVIDIA Jetson.

System Details

  • Hardware: NVIDIA Jetson AGX Orin

  • JetPack Version: 7.2.1

Installation Method:

I am using the following command to install:

curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash

Error Description:

During the bootstrapping process, the installation script throws the following errors (as recorded in nemoclaw.log):

  1. WARNING: Unsupported JetPack version detected. GPU may not be supported

  2. Error: Docker did not accept the Jetson NVIDIA runtime GPU mode for managed bootstrap

This is my very first time using NemoClaw, and I plan to test it using the OpenClaw agent and Ollama with the Nemotron-3 model.

Could you please guide me on how to fix these warnings and errors so that I can successfully proceed with the NemoClaw installation and setup?

Thank you for your support.
nemoclaw.log (292 KB)

2

Hello @Hung_Tran!

Based on the title and content of your topic, it looks like it may receive better visibility and feedback in a different category. We took the liberty of moving it for you.

If this was an incorrect assessment, please send me a direct message.

Disclaimer: this moderation suggestion and message were generated with AI assistance.

5

If cat /etc/docker/daemon.json doesn’t produce following edit it to be identical to:

{
    "default-runtime": "nvidia",
    "runtimes": {
        "nvidia": {
            "args": [],
            "path": "nvidia-container-runtime"
        }
    }
}

Run following to try again. first 3 were extracted from your log.

sudo usermod -aG docker $USER
newgrp docker 

source "${NVM_DIR:-$HOME/.nvm}/nvm.sh" && nvm use 22

curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash -s -- --fresh
6

Thank you very much for moving my topic to the appropriate category.

7

Hi,

WARNING: Unsupported JetPack version detected. GPU may not be supported

This is a harmless warning from ollama.

Error: Docker did not accept the Jetson NVIDIA runtime GPU mode for managed bootstrap

This is a known issue and is fixed around 3 weeks ago.
Could you try it again?

https://github.com/NVIDIA/NemoClaw/issues/11197

Thanks.

8

Thank you for the guidance, I followed it, but the installation process gets stuck or throws warnings during step [6/8] Creating sandbox. I have attached the latest log file (2nd_nemoclaw.log) for your reference. For context, my system is a fresh install using NVIDIA SDK Manager with JetPack 7.2.1 (full options), and I only installed Firefox and jtop prior to running the NemoClaw installer.
2nd_nemoclaw.log (16 KB)

2nd_nemoclaw_screenshort
2nd_nemoclaw_screenshort2494×1408 371 KB

9

Thank you for the guidance. I re-flashed JetPack 7.2.1 and followed the instructions, but I am still encountering an error at step [6/8] Creating sandbox due to a timeout when pulling the Docker image. Could you please advise on how to resolve this?

3rd_nemoclaw
3rd_nemoclaw2560×1440 370 KB

10

You may be better off waiting for Nvidia to reply to you. but you could try

nemoclaw onboard --resume

If it fails again you could try:

sudo apt install binutils 
sudo apt install --reinstall  nvidia-container-toolkit nvidia-container-toolkit-base libnvidia-container-tools libnvidia-container1

CUDA_VER=$(basename "$(realpath /usr/local/cuda)" | sed 's/^cuda-//')
sudo apt install "cuda-compat-${CUDA_VER/./-}"

sudo reboot

# Then

curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash -s -- --fresh
11

Thank you for the guidance and troubleshooting steps provided. I followed them and generated 3rd_nemoclaw.log, but after rebooting and running the installation again, I am still encountering a timeout error during step [6/8] Creating sandbox when pulling the Docker image (spawnSync docker ETIMEDOUT), as shown in 4th_nemoclaw.log.
3rd_nemoclaw.log (13.2 KB)
4th_nemoclaw.log (13.1 KB)
I will have to wait for further guidance from NVIDIA.

12

You could try this to see if it will identify the entire docker image-name:value that nemoclaw attempted to pull in ‘4th nemoclaw.log’

docker images --digests --no-trunc | grep -E 'nemoclaw|openclaw'

If that returns sha256:7dcb60465421... then copy that entire actual value sha256:7dcb60465421??????????????
and

docker pull ghcr.io/nvidia/nemoclaw/openclaw-sandbox@sha256:7dcb60465421-and-the-rest-of-the-value

Then try:

nemoclaw onboard --resume
14

I tried running docker images --digests --no-trunc | grep -E 'nemoclaw|openclaw', but it returned nothing as shown in 5th_nemoclaw.log
5th_nemoclaw.log (423 Bytes)