This repository covers the scripts for running RStudio with Ansible on cloud services. The RStudio Dockerfile is based off Rocker/Tidyverse - which is handy for bioinformatics users, with changes made to the read permissions of RStudio server paths to enable running of RStudio interactively. The container is first built with Docker, then pulled from the Docker daemon to build a Singularity container. This allows the container to be used on HPC if required.
This deployment uses Singularity to build an RStudio Singularity container.
-
It is recommended that users change their Singularity cache directory to a storage volume, as the default is in the
/homedirectory, i.e./home/ubuntu/.singularity/cache, which can run out of space quickly. To change it, simply create a new directory and update the$SINGULARITY_CACHEDIRvariable to it. It is also recommended that you have your storage volume mounted to a directory named/data.Skip this step if you do not have a storage volume.
mkdir /data/singularity_cache SINGULARITY_CACHEDIR=/data/singularity_cache -
Ensure that you have port 8787 open for your instance, you can do so on the Nimbus dashboard following these instructions.
This is an interactive deployment, such that when the Ansible playbook script is run, the user will be prompted to enter any R packages or Bioconductor packages that is required. While libraries can be installed post-container build, it is encouraged to build them into the container for software dependency efficiencies.
Note that the time taken for this deployment varies, and if run for the first time, will require at least 10 minutes for the RStudio container to be built and pulled as a Singularity image. Ensure that your computer does not go to sleep during this time.
This Ansible playbook works on Ubuntu 18.04 and Ubuntu 20.04. Other operating systems and versions will require testing. Raise a ticket if you face issues running this with other operating systems.
Ansible needs to be installed on the machine (instructions provided below).
We have tested and can currently support only R versions 4.1.0, although other versions can still be trialed.
Some R packages and Bioconductor packages require specific libraries on the operating system. Container deployments such as this RStudio only include the most basic default Linux libraries. We have included a few extras that users have brought to our attention that is common for some bioconductor packages. If you require other libraries, feel free to edit the Dockerfile to include them.
sudo apt install --yes software-properties-common
sudo add-apt-repository --yes --update ppa:ansible/ansible
sudo apt install --yes ansible
git clone https://github.com/PawseySC/rstudio-on-nimbus
cd rstudio-on-nimbus
ansible-playbook ansible-rstudio.yaml
#If you make a mistake answering the prompts, cancel (control+c) and rerun the `ansible-playbook` command.
Once you have run the above, you can login to your RStudio session from a web browser using your instance IP address and the port 8787, e.g. 146.118.XX.XX:8787
# You should see (if R=4.1.0):
> .libPaths()
[1] "/home/rstudio/R/x86_64-pc-linux-gnu-library/4.1"
[2] "/usr/local/lib/R/site-library"
[3] "/usr/local/lib/R/library"
From here, you can load the libraries you require, add data to /home/rstudio, and run the analyses you require. Note that /home/rstudio on RStudio corresponds to /data/rstudio on your Nimbus instance.
To end the RStudio session from your instance:
lsof -ti:8787 | xargs kill -9
It is advisable to clean the Singularity cache from time to time - when image pulls are completed. To clean it, first list and check that you are happy to remove the cache:
singularity cache list -v
singularity cache clean
** Use sudo if you did not change the $SINGULARITY_CACHEDIR variable **
sudo singularity cache list -v
sudo singularity cache clean