RStudio Server¶

RStudio Server provides a web-based integrated development environment (IDE) for R, allowing you to run interactive R sessions on Hyperion compute nodes while accessing them through your local web browser. This approach combines the convenience of RStudio's familiar interface with the computational power of the cluster.

Overview¶

When you launch RStudio Server on Hyperion, a dedicated R session runs on a compute node with access to:

• Cluster resources — High-memory nodes, multiple CPU cores, and fast parallel storage
• Your files — Direct access to home, scratch, and project directories
• Installed software — All R packages and modules available on the cluster
• Persistent sessions — Your R environment persists for the duration of your job

The connection workflow is:

┌──────────────┐      SSH Tunnel       ┌──────────────┐      Internal      ┌──────────────┐
│ Your Machine │ ──────────────────────│  Login Node  │ ─────────────────▶ │ Compute Node │
│   Browser    │     (encrypted)       │   hyperion   │                    │   RStudio    │
└──────────────┘                       └──────────────┘                    └──────────────┘

This ensures secure, encrypted access to your session from any location.

Instructions¶

Step 1: Request Compute Resources¶

RStudio Server must run on a compute node, not on the login node. You can request resources either interactively or through a batch job.

Option A: Interactive Session¶

For exploratory work or shorter sessions, request an interactive allocation:

srun -n1 -N1 -c4 --mem=16G --time=04:00:00 -p regular --pty bash

Adjust the resources according to your needs:

Option B: Batch Job¶

For longer sessions or automated workflows, create a batch script named rstudio.slurm:

#!/bin/bash
#SBATCH --job-name=rstudio
#SBATCH --output=rstudio-%j.out
#SBATCH --error=rstudio-%j.err
#SBATCH --time=08:00:00
#SBATCH --cpus-per-task=4
#SBATCH --mem=32G
#SBATCH --partition=regular

module load RStudio-Server
start_rstudio_server.sh

Submit the job:

sbatch rstudio.slurm

Monitor your job status:

squeue -u $USER

Once the job is running, retrieve the connection information from the output file:

cat rstudio-<JOBID>.out

Step 2: Start RStudio Server¶

Once your allocation is active, execute the startup script:

start_rstudio_server.sh

The script will display connection information:

• Interactive session: The information appears directly on screen
• Batch job: The information is written to the output file (rstudio-<JOBID>.out)

================================================================================
                         RSTUDIO SERVER SESSION INFO
================================================================================

  Job ID:       1234567
  Node:         hyperion-107.sw.ehu.es
  Port:         8945
  Username:     username
  Password:     xK9mPq2wL5nR

--------------------------------------------------------------------------------
  CONNECTION INSTRUCTIONS
--------------------------------------------------------------------------------

  1. Open a terminal on your local machine and run:

     ssh -N -L 9999:localhost:8945 -J username@hyperion.sw.ehu.es hyperion-107

  2. Open your web browser and navigate to:

     http://localhost:9999

  3. Log in with the credentials above.

  4. To terminate the session, cancel the Slurm job:

     scancel 1234567

================================================================================

Step 3: Create SSH Tunnel¶

The SSH tunnel securely forwards traffic from your local machine to the compute node. Open a new terminal window on your local machine and run the SSH command provided in the connection information.

The tunnel command follows this pattern:

ssh -N -L <local-port>:localhost:<remote-port> -J <username>@hyperion.sw.ehu.es <node>

Where:

Linux and macOS¶

Open a terminal and run:

ssh -N -L 9999:localhost:8945 -J username@hyperion.sw.ehu.es hyperion-107

Replace username, 8945, and hyperion-107 with your actual values from the connection information. You may also change 9999 to any available port on your local machine.

Windows¶

Using PowerShell or Command Prompt (Windows 10/11)¶

Windows 10 and later include a built-in SSH client. Open PowerShell or Command Prompt and run:

ssh -N -L 9999:localhost:8945 -J username@hyperion.sw.ehu.es hyperion-107

Adjust the values as described above.

Step 4: Access RStudio¶

1. Open your web browser (Chrome, Firefox, Safari, or Edge)
2. Navigate to http://localhost:<local-port> using the local port from your tunnel command (e.g., http://localhost:9999)
3. Enter your cluster username and the password from the connection information
4. The RStudio interface should now appear

Working with RStudio on the Cluster¶

File Access¶

RStudio has direct access to all your cluster files:

Navigate using the Files pane in RStudio or R commands:

# Check current directory
getwd()

# Change to scratch
setwd("/scratch/username")

# List files
list.files()

Using R Modules¶

Before starting RStudio Server, you can load additional R-related modules that will be available in your session:

module load R/4.3.0
module load RStudio-Server
start_rstudio_server.sh

Parallel Computing¶

Your R session has access to the CPU cores you requested:

# Check available cores
parallel::detectCores()

# Example with mclapply
library(parallel)
results <- mclapply(1:100, my_function, mc.cores = 4)

Other parallelization options include foreach/doParallel and future/future.apply.

Memory Management¶

R can only use the memory requested in your Slurm allocation. Monitor usage with gc() and check object sizes with object.size(). If you encounter "cannot allocate vector" errors, submit a new job with higher --mem.

Efficient Data I/O¶

For large datasets, consider using data.table::fread() or vroom::vroom() instead of base R functions. For saving R objects, saveRDS()/readRDS() or the qs package provide fast binary storage.

Ending Your Session¶

Saving Your Work¶

Before ending your session:

1. Save all open files: File → Save All
2. Save your R workspace if needed: save.image("/path/to/workspace.RData")
3. Note any file paths you'll need for future sessions

Terminating the Session¶

You can end your RStudio session in several ways:

From your local machine:

scancel <JOBID>

From the RStudio interface:

• Close the browser tab and let the job expire naturally, or
• Use the terminal within RStudio to run scancel $SLURM_JOB_ID

Important: Simply closing the browser does not terminate the Slurm job. The job continues running until it reaches its time limit or is explicitly cancelled.

Troubleshooting¶

Connection Refused¶

If you see "Connection refused" when accessing your localhost URL:

1. Verify the SSH tunnel is running (check the terminal where you started it)
2. Confirm the port numbers are correct:
   • The local port in your tunnel must match the port in your browser URL
   • The remote port in your tunnel must match the port from connection info
3. Ensure the RStudio Server is still running: squeue -u $USER

Connection Reset¶

If the connection drops immediately:

• Use the ProxyJump (-J) syntax as shown in the examples
• Verify you're connecting to the correct compute node

Authentication Failed¶

If RStudio rejects your credentials:

• Use your cluster username (not email)
• Use the password from the connection info (not your cluster password)
• Passwords are case-sensitive

RStudio Interface Not Loading¶

If the browser shows a blank page or loading indicator:

• Clear browser cache and reload
• Try a different browser
• Check the Slurm error file: cat rstudio-<JOBID>.err

Session Expired¶

If your session becomes unresponsive:

• Your Slurm job may have reached its time limit
• Check job status: squeue -u $USER
• Submit a new job with longer --time if needed

Best Practices¶

1. Request appropriate resources: Don't over-request; it increases queue time
2. Use scratch for large files: Keep your home directory clean
3. Monitor your jobs: Check squeue periodically
4. Clean up: Cancel jobs you no longer need with scancel
5. Use batch jobs for long sessions: They're more reliable than interactive sessions

Getting Help¶

If you encounter issues not covered in this documentation:

• Check the Slurm output and error files for your job
• Review the RStudio Server logs: ~/.local/share/rstudio/log/
• Contact support with your job ID and a description of the problem