ivillasante
commented
10 months ago Rstudio Server in Singularity
A wiki entrance explaining how to download and setup a Singularity container for R dev. (Rstudio server / Shiny server / other R stuff) in the cluster.
It seems that some of the members of the group have had issues with reproducibility working with R and the main idea is to propose a standard protocol to abstract and potentially remove all the complexity of handling system dependencies and package versions with containers. A collateral effect is being able to very easily run rstudio server directly in the cluster. The entrance could have the following main points:
- Where to find pre-built development containers with different flavours (tidyverse, vep, bioconductor, shiny, single-cell, cuda, pyenv and more) rocker project
- How to download the images
- How to modify already existing images to build your own customized versions.
- Where to store those images (maybe a cloud-based solution (aws / google cloud / dockerhub/ github registry) or an inhouse registry that could be installed and kept on a server or unused workstation ) check Singularity Registry Server
- How to download the images with singularity in the cluster (may vary according whatever is chosen on step 4) It would be nice to find a common space in the cluster where to store the cahce maybe something like
/workspace/temp/.cache/containers- How to start the server in the cluster
- How to connect to the server remotely from your local machine. (There are many options for 6 & 7, but interactive + ssh port forward has been tested successfully by myself)
- The part of setting up a service in the cluster is the same as jupyter nb. ssh -> screen -> interactive -> start service
- The part of starting Rstudio involves some steps:
- Downloading the singularity image
- Generating temp and database files,
- Configuring the path to the r packages library (Optional)
- configuring ports and binds in the container
- Setting up the password
- starting rstudio server with singularity exec command
Here is the script I made which can be executed straight away:
The part of connecting to the active service is just a port forward ssh -N -L local_port:bbgnXX_node:remote_port -p 22022 user@bbgcluster
Rstudio Server in Singularity
A wiki entrance explaining how to download and setup a Singularity container for R dev. (Rstudio server / Shiny server / other R stuff) in the cluster.
It seems that some of the members of the group have had issues with reproducibility working with R and the main idea is to propose a standard protocol to abstract and potentially remove all the complexity of handling system dependencies and package versions with containers. A collateral effect is being able to very easily run rstudio server directly in the cluster. The entrance could have the following main points:
/workspace/temp/.cache/containers