For security reasons, Domino Workspace sessions are only accessible on one port. For example, Jupyter typically uses port 8888. When you launch a Jupyter Workspace session, a Domino executor starts the Jupyter server in a Run, and opens port 8888 to serve the Jupyter application to your browser. If you were to try to use the Jupyter terminal to start another application on a different port, it would not be accessible.
However, in some cases you might want to run multiple interactive applications in the same Workspace session. These cases include:
-
Editing and debugging Dash or Flask apps live
-
Using Tensorboard to view progress of a live training job
Domino 3.5+ supports this with Jupyter Server Proxy and JupyterLab.
Prerequisites
-
Python 3+
-
Jupyter Server Proxy
Jupyter Server Proxy is installed by default in the latest Domino Standard Environments. To install it in one of your existing environments, see the following instructions.
If you are not on the recent version of the Domino Standard Environment, you can use the following steps to install Jupyter Server Proxy in your Domino Environment.
-
Add the following lines to your environment’s Dockerfile Instructions.
# Install NodeJS
# You can omit this step if your environment already has NodeJS 6+ installed
RUN curl -sL https://deb.nodesource.com/setup_8.x | bash - && \
apt-get install nodejs -y && \
rm -rf /var/lib/apt/lists/*
# Switch to the latest JupyterLab start script
RUN rm -rf /var/opt/workspaces/Jupyterlab/start.sh && \
cd /var/opt/workspaces/Jupyterlab/ && \
wget https://raw.githubusercontent.com/dominodatalab/workspace-configs/2019q4-v1/Jupyterlab/start.sh && \
chmod 777 /var/opt/workspaces/Jupyterlab/start.sh
# Install and enable jupyter-server-proxy
RUN pip install --upgrade jupyterlab==0.35.4 && \
pip install nbserverproxy jupyter-server-proxy && \
jupyter serverextension enable --py --sys-prefix nbserverproxy && \
jupyter labextension install jupyterlab-server-proxy
-
Update the JupyterLab definition in the Pluggable Workspace Tools section of your environment.
jupyterlab: title: "JupyterLab (Beta)" iconUrl: "https://raw.githubusercontent.com/dominodatalab/workspace-configs/develop/workspace-logos/jupyterlab.svg?sanitize=true" start: [ /var/opt/workspaces/Jupyterlab/start.sh ] httpProxy: internalPath: "/{{#if pathToOpen}}/lab/tree/{{pathToOpen}}{{/if}}" port: 8888 rewrite: false
If you launch a JupyterLab Workspace session in an environment with Jupyter Server Proxy installed, you can start and serve additional applications if they are served on a different port than JupyterLab.
Once an additional application is started, you can access it at the following URI:
<workspace id>-workspace.<host name>/proxy/<port>
Suppose your JupyterLab session is served at:
`https://<YourDominoURL>/workspace?owner=<ownername>&projectName=<projectname>&runId=<runhash>'
If you then use the JupyterLab terminal to start a Dash app on port 8887 for debugging, you could open the Dash app at:
If you instead use the JupyterLab terminal to start a Bokeh app on port 5006, you could open the Bokeh app at:
With this model you can host multiple applications on different ports and expose each through your JupyterLab workspace.
If you edit the source files in JupyterLab after the application is running, you must restart the app in the browser for the edits will take effect.
For environments that have VSCode installed in JupyterLab, you can start a VSCode session from JupyterLab, and then start an App from VSCode. This allows you to debug using VSCode.
A new App or process that you start and open in a separate tab will not have the Domino Workspace with options to stop, sync, commit, or manage your project files. To access this application and manage your changes, you must open the main JupyterLab tab for your Workspace session.