jupyter_remote

Spawn a Jupyter Lab on a remote server and connect to with SSH port forwarding/tunneling it with your local computer.

It is developed and tested on a Mac air laptop running Mac OS Monterey (version 12.2.1).

The script jupyter_remote_connect.sh will spawn a remote Jupyter Lab and connect it to your local browser using SSH. The script is adopted from https://stackoverflow.com/questions/43696291/script-to-run-jupyter-notebooks-from-remote-server

The script assumes that you are using conda on the remote server. It also uses a screen session, so screen must also be installed.

The script is set up to maintain one Jupyter Lab process, and one SSH connection. If one or more of these are already running (e.g., you ran this script previously and did not shut it down), it should use the existing Jupyter Lab and/or SSH. The output to the screen will reflect this.

Prerequisites:

• Conda must be installed on the remote server, and there must be an environment with jupyterlab included.
• Screen is needed on the remote server to spawn the Jupyter Lab in a screen session.
• A web browser on the local machine. I have tested it with Chrome and Firefox.
• SSH is needed.

Instructions:

• Edit the variables in Section. 1 on top. NO SPACES around the equals signs!
  • Edit the server log in information (username and serverIP. DNS domain name should work instead of the IP address).
  • Specify the directory you want to work in on the server (work_dir).
  • Specify the conda environment name or directory.
  • Indicate which browser to use on the LOCAL computer. It must work with "open -a" on the command line.
  • Specify the ports to use on the local and remote machines. They can be the same. Try not to use a port already being used, especially on the remote server.
• Make the script executable, if it is not already. In a terminal, use chmod +x jupyter_remote_connect.sh.
• To run the script, invoke ./jupyter_remote_connect.sh. It may take a few moments for Jupyter Lab to spin up, and meanwhile the browser window may show "site cannot be reached" or similar.
• Shutting down (highly recommended)
  • These instructions are printed out to the screen when you run the script.
  • The easiest way to shut down the Jupyter Lab server and screen session is to use the "File --> Shut Down" menu in the web browser.
  • The tedious way to shut it down on the remote server is to SSH in to the remote server, use the screen -r command to attach to the screen session, then Ctrl-C to kill the Jupyter Lab.
  • To shut down the SSH port forwarding/tunnel connection:
    • Do this on your LOCAL computer
    • ps -u your_local_user_name | grep SSH
    • Look for a line with port forwarding, e.g. -L 8890:localhost:8888 -N.
    • Kill the process with kill pid where pid is the second number in the output from above.