Notebook Session Tips

We've developed these tips to help use notebook sessions.

• We found that our JupyterLab interface runs faster in Firefox compared to other browsers. We noticed that creating notebooks, opening notebooks, writing code, and executing cells run faster in Firefox.

• We recommend that you have no more than one user using a notebook session. Even if that means the notebook session for each user is smaller. We recommend you avoid having many users in the same notebook session because:

  • It causes resource contention. You can run many notebooks all using the same OCPUs and memory. The same applies to dependencies installed by different members of the team that can create conflicts. For example, when using Hadoop.

  • Conflicts when editing and writing the same file.

• Don't run long running Python processes in a notebook cell. The run limit for a notebook cell is about 30 minutes. Run the process directly in the terminal and use Python logging to get updates on the progress of a job. Use nohup or use the UNIX nice command to assign priorities to Python jobs. You could also convert a notebook to a Python script using the JupyterLab conversion tool.

  Try using fewer threads than the number of OCPUs times two.

• To monitor the memory and CPU consumption of workloads, run top or I in a JupyterLab terminal window.

• If you get a JupyterLab Launcher error in a notebook session, then press:

  • On Linux and Windows:

    • Firefox and Chrome: Ctrl + Shift + R

  • On Mac:

    • Firefox and Chrome: Command + Shift + R

    • Safari: Command + Option + R