R & RStudio

Rockfish offers two complementary ways of working with R:

RStudio Server – a full-featured IDE you open in a web browser.
Command-line R – run R scripts or an interactive R prompt from the terminal (or inside Slurm batch jobs).

This page collects all R-related instructions in one place.

1 Interactive RStudio Server 

You can launch RStudio two ways. Both end up running the same server inside the cluster – choose whichever matches your workflow.

1.1 Open OnDemand (graphical, point-and-click)

Connect to the Hopkins VPN – the portal is JHU-internal.
Open your browser and go to https://portal.rockfish.jhu.edu.
Log in with your Rockfish credentials.
On the dashboard click RStudio Server.
Fill in the resource form:
- R version (e.g. R/4.4.1 – RStudio Server)
- R_LIBS_USER – path to your personal R library
- Wall time, cores, partition (see cheatsheet below)
Click Launch. When the session is Running hit Connect to RStudio Server.
A new browser tab opens → happy coding!

1.2 Command-line helper script 

Prefer the shell, need repeatability, or want to embed RStudio in a larger workflow? Use r-studio-server.sh:

Quick start

# 1  SSH to Rockfish
ssh <YourUserID>@login.rockfish.jhu.edu

# 2  Generate a Slurm batch script
r-studio-server.sh -n 1 -c 2 -m 8G -t 4:00:00 -p defq -e <you>@jhu.edu
#             (run r-studio-server.sh -h for all flags)

# 3  Submit the job
sbatch R-Studio-Server.slurm.script

# 4  Read the connection info once the job starts
cat rstudio-server.job.<JOBID>.out

# 5  On *your laptop* create the SSH tunnel it shows
ssh -N -L 8787:node123:8787  <YourUserID>@login.rockfish.jhu.edu

# 6  Open http://localhost:8787  →  RStudio Server

During the first run the helper also writes a hidden config file .r-studio-variables that you may customise.

Note

The script prints a full summary before submitting. Press Ctrl-C to cancel if anything looks wrong.

Parameter reference

Flag	Meaning
`-n`	Nodes (default 1)
`-c`	Cores per task (≈ 4 GB RAM each)
`-m`	Memory – overrides cores × 4 GB if set
`-t`	Wall time `HH:MM:SS` or `D-HH:MM`
`-p`	Partition (`express`, `shared`, `parallel`, `a100`…)
`-q`	QoS (optional)
`-g`	GPUs (GPU partitions only)
`-e`	E-mail for job notifications
`-a`	Account / PI group (required on GPU partitions)

Partition cheatsheet

For available partitions, see: Available Partitions

Helper-script output files

File	Purpose
`R-Studio-Server.slurm.script`	The Slurm batch script that starts the server
`rstudio-server.job.<JOBID>.out`	SSH-tunnel command + URL to open in a browser
`.r-studio-variables`	Hidden config (RStudio & module versions, ports, …)
`~/.config/rstudio/…`	Per-user RStudio settings (created on first launch)

Customising the R version

Edit .r-studio-variables and/or the Slurm script. Load the desired module, e.g.:

module load gfbf/2023b  R/4.4.1-gfbf-2023b
export R_LIBS_USER=${HOME}/rlibs/R-4.4.1-gfbf-2023b

2 Command-line R (outside RStudio)

2.1 Loading R modules 

Rockfish’s default R is currently 4.4.1. Load a different version with the module system:

module avail r            # list every R build
module load  r/3.6.3      # example: older version
module load  r/4.4.1      # example: latest

Some domain-specific software bundles ship their own R:

module load  Seurat/4.1.1   # uses R 4.1.3 internally
module load  edgeR/3.38.1   # uses R 4.2.0

2.2 Installing CRAN / Bioconductor packages 

The simplest way (installs into your personal library):

module load r/4.4.1
R
> install.packages("hdf5r")

Installing from a source tarball:

wget https://cran.r-project.org/src/contrib/LearnBayes_2.15.1.tar.gz
module load r/4.4.1
R CMD INSTALL -l ~/rlibs/R-4.4.1  LearnBayes_2.15.1.tar.gz

Tell R where to look for those packages (add to your .bashrc or to Slurm scripts):

export R_LIBS_USER=~/rlibs/R-4.4.1

2.3 Using Conda 

Conda can manage completely independent R stacks:

module load anaconda
conda create -n hdf5r  r-base=4.4.1  r-hdf5r  -c conda-forge
conda activate hdf5r
R
> q()

You may add further packages later:

conda activate hdf5r
conda install  -c bioconda  bioconductor-limma
conda install  -c conda-forge  r-patchwork

Reminder: activate the env and keep the anaconda module loaded when your Slurm job starts.

2.4 Installing from GitHub (dev tools)

module load r/4.4.1
R
> install.packages("devtools")
> devtools::install_github("davidaknowles/leafcutter/leafcutter",
                           lib = Sys.getenv("R_LIBS_USER"))

2.5 Checking package versions 

packageVersion("leafcutter")

3 Troubleshooting & FAQ 

RStudio job stuck in `PENDING` – the partition is full; pick express or shorten the wall-time.
Cannot open RStudio / Jupyter URL – verify the SSH tunnel on your laptop and that you are browsing http://localhost:<port>.
“Address already in use” – change the local port in the tunnel (e.g. -L 8899:node:8787).
Package installs fail with permissions errors – set R_LIBS_USER to a directory you own (e.g. ~/rlibs/R-4.4.1).
Need GPU-accelerated R – load cuda/a100 modules and see the Rockfish GPU guide for Slurm flags.

Questions? E-mail help@rockfish.jhu.edu.