• What is Docker
• How can Docker be useful to R users
• How Docker works, how to use it
• Examples
• Docker Best Practices

• Containers package a piece of software
• Containers wrap all dependencies for this software to run
• Containers are invisible to each other

• Containers can be combined to create more complex modular systems

• Containers are nice little boxes with reproducibe software inside

• Repeatability
  • Analysis
  • Environment
• Application and data packaging
• Development / Production environment management
• Deployment on well known PaaS
• Easy services deployment

• Dockerfile: describes an image (software, dependencies, …)
• image: the software with all its dependencies
• container: (running) instance of an image

• Choose a base image (for ex. r-base)

  FROM r-base:latest

• Describe your customisations (add R packages, your code, …)

  RUN R -e "install.packages('shiny')"
  COPY build/mypackage.tar.gz .

• Build the image

  $ docker build -t my-image-name .

• Start a container

  $ docker run my-image-name

• Each new container starts from the image, data are lost when the container is deleted

• Bind mounts are filesystem locations outside the container where data can be read or saved

  $ docker run --mount type=bind,source=/path/to/my/data,target=/data image-name

  Docker Documentation: Use bind mounts

• Install Docker

  Docker Install Documentation

• Run an image

  $ docker run --rm -p 8787:8787 -e PASSWORD=GoodPassword
    --mount type=bind,source=~/Documents/Dev/Docker4R,target=/home/rstudio/data
    rocker/tidyverse

    FROM r-base:latest
    MAINTAINER Lynda Metref "lynda.metref@gmail.com"

    RUN R -e "install.packages('shiny')"

    # Expect a single file in the build directory
    COPY build/RBoilerplateShinyDocker*.tar.gz .
    RUN R -e "install.packages(dir(pattern='RBoilerplateShinyDocker.*'))"

    EXPOSE 8888
    ENTRYPOINT ["R", "-e", "library(RBoilerplateShinyDocker); launch(8888,'0.0.0.0')"]

Source

• Don't let strangers in
• Only give your keys to someone you know well

• Docker Hub Certified Images and Verified Publisher
• Reputable communities
• Check Dockerfiles

• Do not run your application in the container as root (the default)

  $ docker run --user my_user ...

• Check network parameters

  Docker Network Documentation

• Don't put credentials in Dockerfile or image, load them from environment.

Sources:

• https://ropenscilabs.github.io/r-docker-tutorial
• https://docs.docker.com

Thank you to:

• Helbling Technik AG for supporting the event organisation
• Etienne Dysli Metref for review and feedback
• R User Group for their invitation