magic4wei/miniconda repository overview

⁠Installed Packages

• magic4wei/miniconda:miniconda3-<XXXXXX>-ubuntu<XX>.04⁠: installed wget, openssl, ca-certificates and Miniconda3.
• magic4wei/miniconda:miniconda3-<XXXXXX>-ubuntu<XX>.04-gui⁠: built from magic4wei/ubuntu:gui-base-*⁠, installed Miniconda.

⁠Usage

⁠Explore Miniconda Base Images

Get started with base image magic4wei/miniconda:miniconda3-<XXXXXX>-ubuntu<XX>.04-gui If you want to run a GUI application that requires OpenGL rendering (see tutorial below). Otherwise magic4wei/miniconda:miniconda3-<XXXXXX>-ubuntu<XX>.04 would be better since it has smaller size.

Explore Miniconda base images by running

# Ubuntu 18.04 for example
docker run -it --rm magic4wei/miniconda:miniconda3-latest-ubuntu18.04  # no OpenGL rendering is supported
docker run -it --rm magic4wei/miniconda:miniconda3-latest-ubuntu18.04-gui  # enable OpenGL rendering, see tutorial below

⁠Work with OpenGL Rendering

Base image magic4wei/miniconda:*-gui have already installed required packages to enable OpenGL rendering in container. Before you go through this tutorial, make sure that

• at least one GPU is mounted on your machine
• GPU driver is installed correctly
• You can run glxgears in terminal without any error (package mesa-utils is needed)

Firstly, you need to run xhost + on your host machine to enable X client display.

For Intel GPU, simply run

docker run -it --rm \
-e DISPLAY=$DISPLAY \
-v /tmp/.X11-unix:/tmp/.X11-unix:rw \
--privileged \
--device=/dev/dri:/dev/dri \
magic4wei/miniconda:miniconda3-latest-ubuntu18.04-gui

For Nvidia GPU, it is a little bit more complicated. We need to connect Nvidia driver volume when docker image runs, and add Nvidia OpenGL library path to LD_LIBRARY_PATH inside the container. If you want to run Nvidia executable tools like nvidia-smi, you also need to add their path to PATH inside the container.

For example, I have installed Nvidia driver nvidia-430 at /usr/lib/nvidia-430 and /usr/lib32/nvidia-430 (32-bit compatible libraries), and all my Nvidia tools are located at /usr/lib/nvidia-430/bin. Two steps to enable OpenGL rendering in container:

1. Run image with Nvidia driver volume bindings

docker run -it --rm \
-e DISPLAY=$DISPLAY \
-v /tmp/.X11-unix:/tmp/.X11-unix:rw \
--privileged \
-v /usr/lib/nvidia-430:/usr/hostLib/lib/nvidia \
-v /usr/lib32/nvidia-430:/usr/hostLib/lib32/nvidia \
magic4wei/miniconda:miniconda3-latest-ubuntu18.04-gui

1. Add Nvidia OpenGL library path in the container

export LD_LIBRARY_PATH=/usr/hostLib/lib/nvidia:/usr/hostLib/lib32/nvidia:${LD_LIBRARY_PATH} # required for OpenGL rendering
export PATH=/usr/hostLib/lib/nvidia/bin:${PATH} # optional, add this if you want to run tools like nvidia-smi

Then you are able to run GUI applications that requires OpenGL rendering in the container. For example,

# Example: Run glxgears (needs `mesa-utils`, already installed in GUI base images)
glxgears