配置Nvidia Docker

本文主要根据Docker — 从入门到实践这个教程进行学习。

docker是一个进行机器人相关开发与测试的理想环境，而且Nvidia也推出了可以利用GPU的docker镜像，具体的安装指令如下：

docker run -it -d -P \
 -v /tmp/.X11-unix:/tmp/.X11-unix \
 -e DISPLAY=$DISPLAY \
 --name robot \
 --mount type=bind,source=/home/robotsky/docker-vol/robot-vol,target=/home/ \
 nvidia/cuda:11.8.0-devel-ubuntu20.04 /bin/bash

为了可以运行可视化的程序，比如rviz，需要在宿主系统（Ubuntu 22.04）中运行xhost +si:localuser:root指令，安装mesa-utils库后，就可以运行glxgears来运行带有GUI界面的程序了。

Update:

经过学习后，又配置了dockerfile来编译镜像，可以替换Isaac Gym里面的dockerfile来编译含有ros的镜像：

FROM nvcr.io/nvidia/pytorch:20.12-py3
ENV DEBIAN_FRONTEND=noninteractive 

# dependencies for gym and ros
#
RUN apt-get update \
    && apt-get install -y --no-install-recommends \
    libxcursor-dev \
    libxrandr-dev \
    libxinerama-dev \
    libxi-dev \
    mesa-common-dev \
    zip \
    unzip \
    make \
    gcc-8 \
    g++-8 \
    vulkan-utils \
    mesa-vulkan-drivers \
    pigz \
    git \
    libegl1 \
    git-lfs

# install ros1 and ros2
RUN apt install curl gnupg2 sudo neofetch -y \
    && echo "deb [arch=$(dpkg --print-architecture)] http://mirrors.tuna.tsinghua.edu.cn/ros2/ubuntu/ $(. /etc/os-release && echo $UBUNTU_CODENAME) main" | tee /etc/apt/sources.list.d/ros2.list > /dev/null \
    && curl -s https://gitee.com/ohhuo/rosdistro/raw/master/ros.asc | apt-key add - \
    && apt update \
    && apt install ros-foxy-desktop python3-argcomplete -y \
    && sh -c '. /etc/lsb-release && echo "deb http://mirrors.tuna.tsinghua.edu.cn/ros/ubuntu/ $(. /etc/os-release && echo $UBUNTU_CODENAME) main" > /etc/apt/sources.list.d/ros1.list' \
    && apt update \
    && apt install ros-noetic-desktop -y \
    && apt clean

# Force gcc 8 to avoid CUDA 10 build issues on newer base OS
RUN update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-8 8 \ 
    && update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-8 8

# WAR for eglReleaseThread shutdown crash in libEGL_mesa.so.0 (ensure it's never detected/loaded)
# Can't remove package libegl-mesa0 directly (because of libegl1 which we need)
RUN rm /usr/lib/x86_64-linux-gnu/libEGL_mesa.so.0 /usr/lib/x86_64-linux-gnu/libEGL_mesa.so.0.0.0 /usr/share/glvnd/egl_vendor.d/50_mesa.json

COPY docker/nvidia_icd.json /usr/share/vulkan/icd.d/nvidia_icd.json
COPY docker/10_nvidia.json /usr/share/glvnd/egl_vendor.d/10_nvidia.json

WORKDIR /opt/isaacgym

RUN useradd --create-home --no-log-init --shell /bin/bash gymuser \
    && adduser gymuser sudo \
    && echo 'gymuser:123456' | chpasswd
USER gymuser

# copy gym repo to docker
COPY --chown=gymuser . .

# install gym modules
ENV PATH="/home/gymuser/.local/bin:$PATH"
RUN mkdir /home/gymuser/robot \
    && cd python \
    && pip install -i https://pypi.tuna.tsinghua.edu.cn/simple -q -e .

ENV NVIDIA_VISIBLE_DEVICES=all NVIDIA_DRIVER_CAPABILITIES=all