dockerfiles

Dockerfiles used at CTMR that are not coupled to a specific project.

Typical Docker run invocations look like this:

docker run <options> ctmrbio/container_name

The following options are often used:

• --rm: remove container after it's closed (you pretty much always want this).
• -i: Interactive use, connect your terminal to a terminal in the container.
• -t: Create a pseudo-TTY so you get nice things such as tab completion etc.
• -u <userid>:<groupid>: Set user id and group id to run as inside the container. Commands often show -u $(id -u):$(id -g), which will call id to automatically get your user and group id. Note that the group id might be incorrect if working in a shared folder. For work in our shared CTMR folders, set the group ID to 1314 (the group id for ctmrbioinfo). NOTE: This is not how to set user and group ID when running RStudio in the ctmrbio/rstudio container! See details below.
• -p <host_port>:<container_port>: Connect a port on the host (e.g. CTMR-NAS) to a port in the container. Often used for connecting to web services inside a container, such as RStudio or Pavian. The <host_port> can be exchanged for pretty much whatever you want above 1024, but the <container_port> usually has to be set to a specific value, depending on what service is running inside the container.
• -v <host_path>:<container_path>: Mount a path on the host machine into the container. The <container_path> doesn't have to exists beforehand. Can be useful to connect e.g. an output directory as -v /path/to/outdir:/outdir.
• -d: Detach. This is used for images that run/host a service that is acessible without a terminal, e.g. when running an RStudio container that you access via your browser using port forwarding into the container. Remember to docker kill <container_name> when you are done working with your container! Otherwise it will keep running in the background (forever).

rstudio -- The standard CTMR R environment

• Based on bioconductor/release_base2
• Working directory is /input
• Installed packages:
  • tidyverse (ggplot2, dplyr, tidyr, readr, purrr, tibble, strigr, etc.)
  • ape
  • beeswarm
  • dada2
  • data.table
  • dunn.test
  • ggpubr
  • PairedData
  • pheatmap
  • vioplot
  • rafalib
  • RColorBrewer
  • vega
• Jupyter
• Basic python3 data analysis stack:
  • scipy
  • numpy
  • pandas
  • xlrd
  • openpyxl
• OpenJDK Java8 JRE, JDK

The RStudio image is a bit special, as it is built on official Bioconductor image that has some unique implementations to simplify the use of RStudio inside the container. To specify user and group ID when running RStudio inside the container you use -e USERID=<userid> and -e GROUPID=<groupid> when launching the container, instead of the normal -u <userid>:<groupid> argument.

Also note that this image requires that you set a password used to login to RStudio inside the container. This is done by setting an environment variable, PASSWORD, when launching the container: -e PASSWORD=<your_password>. The password cannot be rstudio. The image will shut down directly if no password is set. The username used to login to RStudio in the container is rstudio.

Example command:

docker run -d --rm -it -e PASSWORD=ctmrbio -e USERID=$(id -u) -e GROUPID=1314 -p 8787:8787 -v $(pwd):/input ctmrbio/rstudio

Replace ctmrbio in the above command with a password of your choice. The USERID is set using the command id -u, which is replaced by the ID of your current user. The GROUPID variable can also be replaced with a GID of your choice if you want to run it in a folder that is not one of the shared CTMR folders, or on a different system.

Use this image with SSH port forwarding to access the RStudio interface running inside the container. The default hosting port inside the container is 8787.

The tag ctmrbio/rstudio:latest will always refer to the most recent R/Bioconductor version. Version numbering of this container otherwise works like this:

ctmrbio/rstudio:<image_version>_R<R_version>_Bioc<Bioconductor_version>

Jupyter

The rstudio image also comes with Jupyter installed with kernels for Python 3 and R. There is a script installed in the image called start_jupyter that starts a notebook server in the current directory (the default is /input inside the container). Note that the Jupyter session will not be able to access things outside of the directory you mount into /input when launching the container. Also remember to activate a port forward from a port of your choice to the default Jupyter port of 8888 inside the container.

docker run --rm -it -u $(id -u):1314 -p 8888:8888 -v $(pwd):/input ctmrbio/rstudio start_jupyter

Also note that when using Jupyter, you must use -u <userid>:<groupid> when launching the container, otherwise all files and folders created by the Jupyter session will be owned by root.

picrust2

• Based on ubuntu:18.04
• Working directory is /mnt
• Builds and installs:
  • epa-ng
  • gappa
  • picrust2 conda environment

Example command:

docker run --rm -it -u $(id -u):$(id -g) -v $pwd:/mnt ctmrbio/picrust2

NOTE: The image uses an ENTRYPOINT script that is located in /picrust2/activate_picrust2_env.sh that automatically activates the picrust2 conda environment inside the container.

Deprecated images

Notes for older Docker images that are currently considered deprecated.

rstudio_luisa [DEPRECATED]

• Based on bioconductor/release_base2
• Working directory is /home/rstudio
• Installed packages:
  • vegan
  • RColorBrewer
  • vioplot
  • pheatmap

Example command:

docker run -d --rm -it -u $(id -u):$(id -g) -p 8787:8787 -v $pwd:/home/rstudio ctmrbio/rstudio_luisa

Use this image with SSH port forwarding to access the RStudio interface running inside the container. The default hosting port inside the container is 8787.

DEPRECATED - replaced by ctmrbio/rstudio.

rstudio_dada2 [DEPRECATED]

• Based on bioconductor/release_base2
• DADA2 installed from github sources
• Working directory is /home/rstudio

Example command:

docker run --rm -it -u $(id -u):$(id -g) -p 8787:8787 -v $pwd:/home/rstudio ctmrbio/rstudio_dada2

Use this image with SSH port forwarding to access the RStudio interface running inside the container. The default hosting port inside the container is 8787.

DEPRECATED - replaced by ctmrbio/rstudio.