This is a image with all open source tools for: FreeCAD + PPP + occ_faceter + gmsh + openmc workflow

Note: this image is large, about 10 GB

For evaluation of parallel-preprocessor, use the smaller image ppp-centos instead.

OpenMC with MPI is not supported in this docker image, consider using singularity image to support MPI parallelization on HPC platforms.

docker pull qingfengxia/ppp_openmc for jupyter notebook way docker pull qingfengxia/ppp_openmc_ssh for jupyter notebook and ssh ways, it is based on image qingfengxia/ppp_openmc

https://hub.docker.com/repository/docker/qingfengxia/ppp_openmc

The Dockerfile is here:

The docker image is based on jupyter/mini-notebook , but it can be switched to ubuntu:focal as base image, see instruction at the end of this document for switching base image.

There are 3 ways to use these docker images, choose the one you like

docker run --rm -p 8888:8888 ppp_openmc Jupyter can create a terminal in webbrowser, just like ssh terminal.

-p 8888:8888 is for jupyter notebook web access This docker image can be running in super user mode by giving -e GRANT_SUDO=yes --user root which gives passwordless sudo capacity, to install more software.

Since nuclear material (tendl-2019-hdf5) has been built in, there is no need to map host folder to container like this. In case of some docker image is not built with material data, please use volume mapping such as -v /mnt/windata/MyData/openmc_data/tendl-2019-hdf5/:/mat_dir for material cross section data. The host folder containing "cross-section.xml" must be mapped to /mat_dir

For more details on options on jupyter/mini-notebook base docker image, see https://jupyter-docker-stacks.readthedocs.io/en/latest/using/common.html

-e GRANT_SUDO=yes - Instructs the startup script to grant the NB_USER user passwordless sudo capability. You do not need this option to allow the user to conda or pip install additional packages. This option is useful, however, when you wish to give $NB_USER the ability to install OS packages with apt or modify other root-owned files in the container. For this option to take effect, you must run the container with --user root. (The start-notebook.sh script will su $NB_USER after adding $NB_USER to sudoers.) You should only enable sudo if you trust the user or if the container is running on an isolated host.

to get an interactive bash shell, as ubuntu 20.04

docker run -it --rm  -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY -h $HOSTNAME -v $HOME/.Xauthority:/home/jovyan/.Xauthority -v $PWD/:/workspace/  ppp_openmc  bash

NOTE: X11 forward is not working for root user, so do not add this option -e GRANT_SUDO=yes --user root , perhaps caused by hardcoded absolute home path in /home/jovyan/.Xauthority

Local X11 forwarding has been tested to work on Linux, it may works on Windows but with different environment variables passed in. Local X11 forwarding can be quickly tested by

1. xterm is working without any xhost configuration,

2. matplotlib.pyplot is working from terminal, tested by:

python3 -c "import matplotlib.pyplot as plt; plt.plot(); plt.show()"

To understand why those arguments are needed -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY -h $HOSTNAME -v $HOME/.Xauthority:/home/jovyan/.Xauthority See https://medium.com/@l10nn/running-x11-applications-with-docker-75133178d090

SSH login is enabled by topping up the image with another thin layer, defined in Dockerfile_ssh

SSH has also X11 forwarding turned on after start the docker container: docker run --rm -p 2222:22 -p 8888:8888 -e GRANT_SUDO=yes --user root -it ppp_openmc_ssh bash then start the ssh server by sudo service ssh start

This argument -p 2222:22 map container's port 22 to host 2222 port. To access this container by ssh (-v means verbose, -Y means X11 forwarding): either ssh -v -Y localhost -p 2222 or ssh -v -Y <container_ip> -p 22

see also How to setup an ssh server within a docker container rebuild the image will change ssh server public key, as the ssh server installation will generate new key pair each time.

On Ubuntu, install the sshpass package, then use it like this: sshpass -p 'YourPassword' ssh user@host

Automatically start the ssh server is possible by CMD ["/usr/sbin/sshd","-D"] in Dockerfile_ssh, if the docker container is running non-interactively (without -it option)

In Dockerfile_ssh, the line CMD ["/usr/sbin/sshd","-D"] will override the jupyter’s start_notebook CMD.

Multiple CMD commands: In principle, there should only be one CMD command in your Dockerfile. When CMD is used multiple times, only the last instance is executed. https://www.educative.io/edpresso/what-is-the-cmd-command-in-docker

1. change the base image in Dockerfile before build: FROM ubuntu:focal or docker build with arg lie this --build-arg IMAGE_NAME="qingfengxia/openmc_mpi" \

2. then build with command line, with different ARG values,

sudo docker build -f Dockerfile_ppp_openmc -t ppp_openmc . --no-cache --build-arg NB_USER=jovyan

Note: NB_USER=jovyan is needed to be compatible with jupyter-notebook base image

see gull example in build_openmc_mpi.sh

Singularity will bind host HOME into the image, so python package installed into normal user home's .local will not visible. One solution will be install all python module as root user into systemwide site-packages.

MPI version can be selected during image building, therefore to match with HPC cluster's MPI version.

--build-arg MPI_VER=3.0 \
--build-arg MPI_VERSION=3.0.1 \

This image can then be converted into singularity image and bind the cluster host MPI library.

singularity pull docker://qingfengxia/openmc_mpi
# depends on singularity version, for version 2.x there is no needs to prefix mpirun inside container
mpirun -np 2 singularity exec --bind  ./openmc_mpi.simg openmc -h