Using RStudio or Jupyter Notebook on Roihu

Page undergoing cleanup after Puhti and Mahti retirement

This page contains outdated references to Puhti and/or Mahti, which have now been decommissioned. The page is undergoing changes to reflect CSC's new supercomputer Roihu.

Learn more about Roihu.

RStudio and Jupyter notebooks are convenient options for developing and running R or Python code. The R or Python code is run on a compute node within an interactive session, but the tools themselves are used via a local web browser.

There are two ways to use RStudio Server or Jupyter Notebook.

The first (and easiest) option is to use the Roihu web interface.
The second option is to create a SSH tunnel from a local PC to a compute node. As compute nodes are inaccessible via the Internet, the tunnel needs to go through a login node. This is not possible with Windows PowerShell (it does not support jump servers), and therefore it is not suitable for RStudio or Jupyter Notebook in Roihu. SSH tunneling requires that you have set up SSH keys.

With Linux, macOS and MobaXterm the SSH tunnelling works by default. PuTTy requires filling in the settings to PuTTy tabs, so it is slower and more complicated, but possible.

Workflow for using RStudio or Jupyter Notebook on Roihu

Using the Roihu Web Interface

For instructions on this, see the web interface documentation.

Using SSH tunneling

Note

SSH tunnelling is not yet enabled on Roihu. The instructions below refer to using it on Puhti.

Start interactive session
Load suitable modules and start RStudio or Jupyter Notebook server
Create SSH tunnel from your PC to Puhti compute node
Open RStudio or Jupyter Notebook in local web browser

More detailed information on the different steps to launch RStudio or Jupyter Notebook using SSH tunneling is provided in the following.

Instructions for SSH tunneling

1. Start interactive session

Start an interactive session, for example with sinteractive -i. For more options and maximum limits see the interactive usage page.

2. Load module and start RStudio or Jupyter Notebook server

In the interactive session run:

RStudio

module load r-env
start-rstudio-server

This set up works with any r-env module. It is also possible to launch a multi-threaded RStudio session using start-rstudio-server-multithread, if you have specified multiple cores when starting an interactive session. Details on using threading with R can be found in the r-env documentation.

Jupyter

Jupyter notebooks can be used with the traditional Jupyter Notebook or the more advanced JupyterLab server.

Jupyter Notebook:

module load python-data 
start-jupyter-server

JupyterLab:

module load python-data 
start-jupyterlab-server

It is also possible to use some other pre-installed Python module than python-data, if it includes Jupyter.

This will start the RStudio, Jupyter Notebook or JupyterLab server on the compute node and print out the instructions for next steps.

Keep this terminal open as long as you are working, to keep the RStudio or Jupyter Notebook server running.

3. Create SSH tunnel from your PC to Puhti compute node

With Linux, MacOS and MobaXterm, open a second SSH terminal on your local machine (do not connect to Puhti yet) and run the SSH tunnelling command printed out by the RStudio or Jupyter Notebook start-up script. For example ssh -N -L 8787:localhost:42896 -J john@puhti.csc.fi john@r07c49.bullx
With PuTTy see the SSH tunnelling with PuTTy instructions.
The command does not print out anything specific, unless you get some error, you likely succeeded.

Also keep this terminal open for as long as you are working to have remote access to RStudio.

4. Open RStudio or Jupyter Notebook in local web browser

Open a web browser on your local machine and copy the URL printed out by the start-up script. For example: http://localhost:8787/
For RStudio, insert the credentials printed out by the start-up script.

5. Close the session

Once you have finished:

Exit RStudio or Jupyter Notebook server by entering Ctrl + C in the interactive terminal session on Puhti.
Close (exit) also the interactive session.
Close SSH tunnel with Ctrl + C.

SSH tunnelling with PuTTy

Both RStudio and Jupyter Notebook also print out PuTTy instructions that have to be copied to PuTTy settings. The port numbers and compute node name may change from session to session.

    PuTTy:
    ssh -N -L 8889:localhost:8889 john@r07c49.bullx
    Set Source (8889) and Destination (localhost:8889) in:
    PuTTy -> Connection -> SSH -> Tunnels

Set up SSH tunneling to a login node with PuTTy. Add port forwarding in PuTTy -> Connection -> SSH -> Tunnels :
- Source port: 8889.
- Destination: localhost:8889
- Keep the type as 'Local'.
- Click 'Add'.
Set up SSH tunneling from login node to compute node in PuTTy -> Connection -> SSH:
- Remote command: ssh -N -L 8889:localhost:8889 john@r07c49.bullx 3: Open to start connection.