DGX Spark (GB10) Ubuntu 24.04 下 NemoClaw 部署：Docker Hub 镜像拉取 node:22-slim 超时

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10
, , ,
1

【Environment】

  • Hardware: NVIDIA DGX Spark (GB10)
  • OS: Ubuntu 24.04 LTS
  • Scenario: Setting up the NVIDIA NemoClaw development environment
  • Tools: Docker 27.0.3

【Problem Description】
When running the NemoClaw deployment script, the command `docker pull node:22-slim` keeps timing out and fails, which prevents the Sandbox environment from being created.

【Error Log】
ERROR: Get “https://registry-1.docker.io/v2/”: dial tcp 54.175.xxx.xxx:443: i/o timeout

【Troubleshooting Performed】

  • Basic public network connectivity works fine (ping/curl to public IPs/domains succeed)
  • Tried switching to domestic Docker mirrors, configuring proxy, and disabling the firewall, none resolved the issue
  • Only Docker Hub image pulls are failing; NVIDIA official images can be pulled without problems

【Help Needed】

  1. Solutions for Docker Hub pull timeout on DGX Spark
  2. Offline images or China-accessible mirror addresses required for NemoClaw deployment

Thank you for your support!

【环境】

  • 硬件：NVIDIA DGX Spark (GB10)
  • 系统：Ubuntu 24.04 LTS
  • 场景：部署 NVIDIA NemoClaw 开发环境
  • 工具：Docker 27.0.3

【问题现象】
执行 NemoClaw 部署脚本时，触发 `docker pull node:22-slim`，出现持续超时错误，无法完成镜像拉取，导致 Sandbox 环境创建失败。

【关键日志】
ERROR: Get “https://registry-1.docker.io/v2/”: dial tcp 54.175.xxx.xxx:443: i/o timeout

【已排查】

  • 公网连通性正常（ping/curl 公网IP/域名均成功）
  • 更换国内 Docker 镜像源、配置代理、关闭防火墙均无效
  • 仅 Docker Hub 镜像拉取失败，NVIDIA 官方镜像可正常拉取

【求助需求】

  1. DGX Spark 环境下 Docker Hub 镜像拉取超时的解决方案
  2. NemoClaw 部署所需的离线镜像或国内可用镜像地址