Using mBuild with Docker
Docker and other containerization technologies allow entire applications and their dependencies to be packaged and distributed as images. This simplifies the installation process for the user and substantially reduces platform dependence (e.g., different compiler versions, libraries, etc). This section is a how-to guide for using mBuild with docker.
Prerequisites
A docker installation on your machine. This Docker installation documentation has instructions to get docker running on your machine. If you are not familiar with docker, the Internet is full of good tutorials like these from Docker curriculum and YouTube.
Jupyter Quick Start
After you have a working docker installation, use the following command to start a Jupyter notebook with mBuild and all the required dependencies:
$ docker pull mosdef/mbuild:latest
$ docker run -it --name mbuild -p 8888:8888 mosdef/mbuild:latest
If no command is provided to the container (as above), the container starts a
jupyter-notebook
at the (container) location /home/anaconda/data
.
To access the notebook, paste the notebook URL into a web browser on your
computer. When you are finished, you can use control-C to exit the notebook
as usual. The docker container will exit upon notebook shutdown.
Warning
Containers by nature are ephemeral, so filesystem changes (e.g., adding a new notebook) only persists until the end of the container’s lifecycle. If the container is removed, any changes or code additions will not persist. See the section below for persistent data.
Note
The -it
flags connect your keyboard to the terminal running in the
container. You may run the prior command without those flags, but be
aware that the container will not respond to any keyboard input. In
that case, you would need to use the docker ps
and docker kill
commands to shut down the container.
Persisting User Volumes
If you are using mBuild from a docker container and need access to data
on your local machine or you wish to save files generated in the container,
you can mount user volumes in the container. User volumes will provide a way
to persist filesystem changes made to a container regardless of the container
lifecycle. For example, you might want to create a directory called
mbuild-notebooks
on your local system, which will store all of your mBuild
notebooks/code. In order to make that accessible from within the container
(where the notebooks will be created/edited), use the following steps:
$ mkdir mbuild-notebooks
$ cd mbuild-notebooks/
$ docker run -it --name mbuild --mount type=bind,source=$(pwd),target=/home/anaconda/data -p 8888:8888 mosdef/mbuild:latest
You can easily mount a different directory from your local machine by changing
source=$(pwd)
to source=/path/to/my/favorite/directory
.
Note
The --mount
flag mounts a volume into the docker container. Here we
use a bind
mount to bind the current directory on our local filesystem
to the /home/anaconda/data
location in the container. The files you see
in the jupyter-notebook
browser window are those that exist on your
local machine.
Warning
If you are using the container with jupyter notebooks you should use
the /home/anaconda/data
location as the mount point inside the container;
this is the default notebook directory.
Running Python scripts in the container
Jupyter notebooks are a great way to explore new software and prototype
code. However, when it comes time for production science, it is often
better to work with python scripts. In order to execute a python script
(example.py
) that exists in the current working directory of your
local machine, run:
$ docker run --mount type=bind,source=$(pwd),target=/home/anaconda/data mosdef/mbuild:latest "python data/test.py"
Note that once again we are bind
mounting the current working directory
to /home/anaconda/data
. The command we pass to the container is
python data/test.py
. Note the prefix data/
to the script; this is because
we enter the container in the home folder (/home/anaconda
), but our script
is located under /home/anaconda/data
.
Warning
Do not bind mount to target=/home/anaconda
. This will cause errors.
If you don’t require a Jupyter notebook, but just want a Python interpreter, you can run:
$ docker run --mount type=bind,source=$(pwd),target=/home/anaconda/data mosdef/mbuild:latest python
If you don’t need access to any local data, you can of course drop the
--mount
command:
$ docker run mosdef/mbuild:latest python
Different mBuild versions
Instead of using latest
, you can use the image mosdef/mbuild:stable
for most recent stable release of mBuild.
Cleaning Up
You can remove the container by using the following command.
$ docker container rm mbuild
The image will still exist on your machine. See the tutorials at the top of this page for more information.
Warning
You will not be able to start a second container with the same name (e.g., mbuild), until the first container has been removed.
Note
You do not need to name the container mbuild as shown in the above
examples (--name mbuild
). Docker will give each container a name
automatically. To see all the containers on your machine, run
docker ps -a
.