Purpose

There can be some difficulty in running rstudio server on the HTCF -- one has to do with the compiled software paths that R relies on, another with controlling where R puts downloaded packages. Using a container addresses these issues.

The container is configured so that you can mount a path into the container where any packages installed during an interactive session are saved. If you mount the same path from session to session, then you have a persistent software library. It also comes with most commonly used R software from CRAN and bioconductor pre-installed.

Cite

This is an exact copy of David Tang's post here:

https://davetang.org/muse/2021/04/24/running-rstudio-server-with-docker/

instructions below are adapted for singularity

Dockerfile

The dockerfile in this project loads the ubuntu system dependencies from Dave Tang's blog. I added R CRAN and Bioconductor dependencies. But, what is important is that you will be binding a system directory to the container so that any packages installed in the container will persist outside of it from session to session.

Build and push

docker build -t <dockerhub username>/<image name> .

docker push <dockerhub username>/<image name>

Singularity pull on the cluster

interactive

eval $(spack load --sh [email protected])

# you can fill in your own username/image name here, or feel free to 
# use mine
singularity pull docker://cmatkhan/htcf_rstudio_server

submit via SBATCH

Note: Even afer the job launches and you get output in the SLURM log, you may have to wait a moment for the container to start.

see the script in this project rstudio_singularity.sbatch

There are two cmd line arguments.

$1: path to the singularity image file

$2: path to the persistent library path. This will get bound to /package in the container and it is what will be the first path in the R .libPaths(). If you install a package while using rstudio in the container, it will be saved on the system (so it is persistent -- you won't have to reinstall packages every time you launch the container if you always provide the same path)

example submission

sbatch --mem-per-cpu=10G -N 1 -n 5 --time=120 scripts/rstudio_singularity.sbatch software/htcf_rstudio_server_latest.sif $PWD/R/4.2/