Integrating Posit Workbench with Jupyter Notebook, Launcher, and Slurm

Workbench | Advanced

These steps describe how to integrate Posit Workbench with Jupyter Notebook running with Posit Launcher and Slurm.

Requirements

This integration is intended to be performed on top of an installation of Workbench that has already been configured with Launcher and Slurm.

Step 1. Ensure that Python is available on each node

  • Ensure that Python is available on each node in the Slurm cluster.
    • If needed, you can install Python, pip, and virtualenv on each node following the steps to Install Python.

      Note

      Our recommended installation instructions for Python allow you to make multiple versions of Python available and avoid replacing existing versions of Python when updating system packages.

Step 2. Install Jupyter Notebook, JupyterLab, and Python packages on each node

  • From the previous step, you should still have the PYTHON_VERSION environment variable defined with the version of Python that you installed.

    • If not, you can define this environment variable before you proceed by running the following command and replacing 3.9.14 with the version of Python that you are using:

      Terminal
      $ export PYTHON_VERSION=3.9.14
  • Install Jupyter Notebook, JupyterLab, and the notebook extensions for Workbench and Connect:

    Terminal
    $ sudo /opt/python/${PYTHON_VERSION}/bin/pip install jupyter jupyterlab rsp_jupyter rsconnect_jupyter workbench_jupyterlab
  • Install and enable the Jupyter Notebook extensions:

    Terminal
    $ sudo /opt/python/${PYTHON_VERSION}/bin/jupyter-nbextension install --sys-prefix --py rsp_jupyter   
    $ sudo /opt/python/${PYTHON_VERSION}/bin/jupyter-nbextension enable --sys-prefix --py rsp_jupyter   
    $ sudo /opt/python/${PYTHON_VERSION}/bin/jupyter-nbextension install --sys-prefix --py rsconnect_jupyter   
    $ sudo /opt/python/${PYTHON_VERSION}/bin/jupyter-nbextension enable --sys-prefix --py rsconnect_jupyter   
    $ sudo /opt/python/${PYTHON_VERSION}/bin/jupyter-serverextension enable --sys-prefix --py rsconnect_jupyter
  • (Optional) Install supplemental Python packages:

    Terminal
    $ sudo /opt/python/${PYTHON_VERSION}/bin/pip install altair beautifulsoup4 \
      cloudpickle cython dask gensim keras matplotlib nltk numpy pandas pillow \
      pyarrow requests scipy scikit-image scikit-learn scrapy seaborn spacy \
      sqlalchemy statsmodels tensorflow xgboost

Step 3. Configure Launcher with Jupyter Notebook

  • Add the following lines to the Launcher Jupyter configuration file:

    /etc/rstudio/jupyter.conf
    jupyter-exe=/opt/python/3.9.14/bin/jupyter
    notebooks-enabled=1
    labs-enabled=1
    default-session-cluster=Local
  • If you installed a version other than Python 3.9.14, then you can replace 3.9.14 in the above jupyter-exe setting with the the version of Python that you installed.

Step 4. Restart Workbench and Launcher Services

  • Restart the services:

    Terminal
    $ sudo rstudio-server restart   
    $ sudo rstudio-launcher restart

Step 5. Test Workbench with Launcher and Jupyter Notebooks

  • From your browser, navigate to the Workbench interface and log in.

  • Select New Session and do the following:

    1. Give your session a name.

    2. In the Editor field, select either Jupyter Notebooks or JupyterLab as the IDE.

    3. Click Start Session.

    A screenshot of the New Session dialog on the Workbench homepage, configured to launch a new Jupyter Notebook session locally.

Now, you can use the Jupyter Notebook or JupyterLab interfaces.

Warning

Some local Jupyter configurations may prevent the JupyterLab session from correctly launching in Workbench. For example, setting a password in the files ~/.jupyter/jupyter-server-config.json or ~/jupyter/jupyter-server-config.py, will cause the JupyterLab session to start but not load through the Workbench interface. Commenting out the configuration in question is sufficient to restore expected functionality.

(Optional) Configure multiple Python versions or environments

The Python integration steps described above result in a single Python environment that contains both core packages for Jupyter Notebook, as well as Python packages for end users.

While this is a simple approach, this setup can result in issues if end users want to use different versions of the same package or if some packages conflict with core packages for Jupyter Notebook.

If you would like to use multiple versions of Python or different Python environments, or if you want to install Jupyter Notebook in a separate environment from Python packages for end users, then you can refer to the documentation for using multiple Python versions and environments with Jupyter.

Additional Documentation

For more information on Workbench and Launcher, refer to the following reference documentation:

Back to top