Skip to content

Interactive Apps

Interactive apps let you run graphical and interactive applications on the compute nodes of the cluster, directly from your browser. Examples include a full remote desktop and programming environments such as JupyterLab or RStudio. Each session runs as a regular Slurm job: the portal submits the job for you, waits for it to start, and then lets you connect to the application.

Currently available apps are:

  • Jupyter: JupyterLab, a web-based environment for interactive notebooks that combine code, results, plots, and text.
  • Code Server: Visual Studio Code running in your browser.
  • Marimo: A reactive Python notebook.
  • RStudio Server: The RStudio development environment for R, running in your browser.
  • Desktop: A full graphical Linux desktop running on a compute node, displayed in your browser. Use it to run graphical applications that have no web interface.

All interactive apps work the same way. This page describes the general steps, and the pages dedicated to each app describe their specific options.

Launching a session

  1. Select Interactive Apps in the top menu of the portal, then choose the application you want to run.

  2. Fill in the form with the resources you need for your session. The exact fields depend on the app, but they usually include:

    • Number of cores and Memory: the CPU cores and memory reserved for the session.
    • Walltime: the maximum duration of the session. The session is stopped automatically when this limit is reached.

  3. Click Launch. The portal submits the job and takes you to the My Interactive Sessions page.

The portal remembers the values you entered, so the form is already filled in the next time you launch the same app.

Request only what you need

Like any other job, a session waits in the queue until the requested resources become available. Smaller requests usually start faster.

Following the state of a session

On the My Interactive Sessions page, each session is shown as a card. The badge in the top right corner of the card shows the state of the session:

  • Queued: the job is waiting in the queue for resources to become available.
  • Starting: the job has started, and the application is being launched on the compute node.
  • Running: the application is ready, and you can connect to it.
  • Completed: the session has ended, either because you closed it or because its time limit was reached.

While the session is running, the card also shows the compute node the session runs on, the time remaining before it is stopped, and a link to the session ID. This link opens the session's directory in the File Browser, where the output.log file contains the output of the application. Check this file first if a session fails to start or behaves unexpectedly.

Connecting to the application

Once the session is Running, a button to launch the application appears on the card, for example Connect to Jupyter or Launch Desktop. Click it to open the application in a new browser tab.

You can close the browser tab and connect again later from the My Interactive Sessions page: the session keeps running on the cluster until it is closed or reaches its time limit. This also means you can start a long computation, close your browser, and come back to check on it later.

Ending a session

When you are done, click Delete on the session card to end the session and release its resources. Closing the browser tab is not enough: the session keeps running, and keeps reserving its resources, until you delete it or its time limit is reached. The session card usually explain how to exit a session cleanly.

Save your work before the time limit

When a session reaches its time limit, it is stopped immediately, and any unsaved work in the application is lost. Keep an eye on the remaining time shown on the session card, and save your work regularly.