Skip to content

Notebook Sessions

A Notebook Session provides a browser-based Jupyter environment for interactive analysis. It is a good place to explore data, test a reduction, and prepare a command for a headless Session.

Launch a Notebook

  1. Sign in to the Science Portal.
  2. Under Launch New Session, choose Notebook and an image that contains your analysis software. Complete the resource choices offered by the form and launch the Session.
  3. Find the new entry under Active Sessions. Wait for it to start, then open its notebook link to enter Jupyter.

Available packages, graphics processors (GPUs), and resource limits depend on the selected image and server. If the Session does not start, check its status in the portal and consult Support.

Create, run, and save a notebook

  1. In Jupyter's file browser, navigate to /arc/home/USER, replacing USER with your CADC username. You can use /arc/projects/PROJECT instead if your project has given you write access.
  2. Open a launcher and choose a Python notebook, or use New and select a Python kernel if your image provides the classic notebook interface.

  3. Enter the following code in the first cell, then press Shift+Enter:

    print("Ready to analyse your data")

  4. Confirm that the message appears below the cell. Rename the notebook to first-analysis.ipynb and choose File > Save Notebook in JupyterLab, or Save and Checkpoint in the classic interface.

  5. Confirm that the saved file appears under your chosen /arc directory. Use Storage Management to upload inputs or download a copy of the notebook.

Save your results before you finish. Return to Active Sessions in the Science Portal, use the delete control for this Session, and confirm the deletion. Files saved under /arc remain; /scratch and unsaved kernel state do not. Closing Jupyter's browser tab does not delete the Session.

Launch from the command line (optional)

After installing the client, list the images available on the active server before choosing one:

canfar login cadc
canfar image ls --kind notebook
canfar create notebook IMAGE_NAME --name analysis

Use canfar create --help to check the resource options supported by your installed client.

Monitor the Session and open it when it is ready:

canfar ps --all
canfar info SESSION_ID
canfar open SESSION_ID

If it remains Pending, inspect canfar events SESSION_ID and the batch troubleshooting guide. A Session that is not ready has no application logs yet; use canfar logs after it has started.

Work with files

Use the mounted paths supplied by your server:

/arc/home/<user>/          personal persistent files
/arc/projects/<project>/   project files, when available
/scratch/                  temporary Session-local staging

Save notebooks, code, and results under /arc or a persistent VOSpace Service. /scratch is useful for high-I/O intermediates and is deleted when the Session ends. For remote data that is not mounted, use CANFAR data transfers or the Python filesystem helpers.

The Jupyter file browser and terminal operate inside the Session. Large local uploads are often more reliable when copied explicitly with canfar data cp.

Use a specialised image

If a published image includes CASA or another astronomy package, follow that image's documentation. Do not assume that every Notebook has the same Python packages or that installing a package in one Session changes another. For repeatable work, pin dependencies in a Container Image or versioned environment.

Some deployments also expose shared software through CVMFS. Treat it as read-only and record the module and version used by a workflow.

Move from exploration to automation

Keep the notebook for inspection and use a script with explicit input/output paths for repeatable reductions:

from pathlib import Path

input_path = Path("/arc/projects/<project>/input.fits")
output_path = Path("/arc/projects/<project>/results/output.fits")
# Load input_path, run the reduction, and write output_path.

Then submit the same command as a headless Session, following the batch guide. Record the image, resource request, Storage Identifiers, and code revision with the run.

Troubleshooting

  • A kernel that will not start may indicate an image or resource problem; inspect info, events, and the Session resource request.
  • A missing file is usually a path, identifier, or group-permission issue; confirm it with canfar data info IDENTIFIER:/path.
  • A slow notebook may be reading a remote object repeatedly. Stage it once to /scratch or use the opt-in cache guidance in Filesystem and Python tools.
  • A lost browser connection does not necessarily mean the Session stopped; check canfar ps --all first.

See Support when the checks do not identify the cause.