如何解决Docker容器中Nvidia Nsight Systems图形界面启动失败问题

我需要创建一个能运行Nvidia Nsight Systems图形界面的Docker容器，用来生成CUDA和Python应用的性能报告。构建容器过程顺利，但执行nsys-ui启动图形界面时出现如下错误：

################# ERROR: CrashReporter #################
CrashReporter

Qt initialization failed

Failed to load Qt platform plugin: "xcb"
Library path: /opt/nvidia/nsight-systems/2023.4.1/host-linux-x64/Plugins/platforms/ http://libqxcb.so
Application is going to abort
sh: 1: xmessage: not found
/usr/local/bin/nsys-ui: line 63: 38 Aborted (core dumped) "$NV_AGORA_PATH/CrashReporter" --hide-stack "NVIDIA Nsight Systems" "NsightSystems" "2023.4.1 (Build 2023.4.1.97-234133557503v0)" "$NV_QUADD_PATH" "$@"
""

我尝试在Dockerfile中添加libxcb-xinerama0、libxcb-cursor0等依赖库，但问题仍未解决。以下是我使用的Dockerfile、构建及运行命令，求可行的解决方案或完整Docker配置方案。

Dockerfile内容 #

FROM nvcr.io/nvidia/cuda:12.0.1-devel-ubuntu22.04

ENV DEBIAN_FRONTEND noninteractive

RUN apt-get update -y && \
    apt-get install -y --no-install-recommends \
        apt-transport-https \
        ca-certificates \
        dbus \
        fontconfig \
        gnupg \
        libasound2 \
        libfreetype6 \
        libglib2.0-0 \
        libnss3 \
        libsqlite3-0 \
        libx11-xcb1 \
        libxcb-glx0 \
        libxcb-xkb1 \
        libxcomposite1 \
        libxcursor1 \
        libxcb-xinerama0 \
        libxdamage1 \
        libxi6 \
        libxml2 \
        libxrandr2 \
        libxrender1 \
        libxtst6 \
        libgl1-mesa-glx \
        libxkbfile-dev \
        openssh-client \
        libxcb-cursor0\
        wget \
        xcb \
        xkb-data && \
    apt-get clean

RUN cd /tmp && \
    wget https://developer.nvidia.com/downloads/assets/tools/secure/nsight-systems/2023_4_1_97/nsight-systems-2023.4.1_2023.4.1.97-1_amd64.deb && \
    apt-get install -y ./nsight-systems-2023.4.1_2023.4.1.97-1_amd64.deb && \
    rm -rf /tmp/*

# CMD ["nsys-ui"]

构建命令 #

docker build -f nsight-systems.Dockerfile --no-cache --tag=nsight-systems:2023.4 .

运行命令 #

docker run -it --rm --gpus all -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix --cap-add=SYS_ADMIN --security-opt seccomp=unconfined -v $(pwd):/mnt --network=host nsight-systems:2023.4

启动命令 #

nsys-ui

1. 补充缺失的Qt xcb依赖库 #

你的Dockerfile中仍缺少部分xcb相关依赖，同时错误提示xmessage: not found，需添加以下包：

libxcb-util1 libxcb-image0 libxcb-keysyms1 libxcb-render-util0 libxcb-shape0 libxcb-sync1 libxcb-xfixes0 libxcb-xinput0 x11-utils

2. 更新后的Dockerfile #

FROM nvcr.io/nvidia/cuda:12.0.1-devel-ubuntu22.04

ENV DEBIAN_FRONTEND noninteractive
# 明确指定Qt平台插件路径
ENV QT_QPA_PLATFORM_PLUGIN_PATH=/opt/nvidia/nsight-systems/2023.4.1/host-linux-x64/Plugins/platforms

RUN apt-get update -y && \
    apt-get install -y --no-install-recommends \
        apt-transport-https \
        ca-certificates \
        dbus \
        fontconfig \
        gnupg \
        libasound2 \
        libfreetype6 \
        libglib2.0-0 \
        libnss3 \
        libsqlite3-0 \
        libx11-xcb1 \
        libxcb-glx0 \
        libxcb-xkb1 \
        libxcomposite1 \
        libxcursor1 \
        libxcb-xinerama0 \
        libxdamage1 \
        libxi6 \
        libxml2 \
        libxrandr2 \
        libxrender1 \
        libxtst6 \
        libgl1-mesa-glx \
        libxkbfile-dev \
        openssh-client \
        libxcb-cursor0 \
        wget \
        xcb \
        xkb-data \
        # 新增缺失依赖
        libxcb-util1 \
        libxcb-image0 \
        libxcb-keysyms1 \
        libxcb-render-util0 \
        libxcb-shape0 \
        libxcb-sync1 \
        libxcb-xfixes0 \
        libxcb-xinput0 \
        x11-utils && \
    apt-get clean && \
    rm -rf /var/lib/apt/lists/*

RUN cd /tmp && \
    wget https://developer.nvidia.com/downloads/assets/tools/secure/nsight-systems/2023_4_1_97/nsight-systems-2023.4.1_2023.4.1.97-1_amd64.deb && \
    apt-get install -y ./nsight-systems-2023.4.1_2023.4.1.97-1_amd64.deb && \
    rm -rf /tmp/*

3. 简化运行命令 #

无需额外的权限配置（除非有特殊需求），运行命令可简化为：

docker run -it --rm --gpus all -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v $(pwd):/mnt nsight-systems:2023.4

4. 主机端X权限配置 #

确保主机允许容器访问X服务器，执行：

xhost +local:root

（注：此为临时开放权限，若需永久配置可修改/etc/X11/Xwrapper.config或设置更严格的xhost规则）

验证启动 #

进入容器后执行nsys-ui，图形界面应可正常启动。

内容的提问来源于stack exchange，提问作者Gota_12