Useful bash scripts

JupyterLab server on cluster via SSH tunnel

Tested on UChicago RCC.

1. Edit jupyterlab.sbatch to update the JupyterLab session resource. Current setup is that the server will be created on partition broadwl, 1 node, 1 CPU, 2G memory and is good for 24 hours.
2. Submit the job script from the login node

sbatch jupyterlab.sbatch

3. Wait for the job to get started. That is, you should find a non-empty nb-log-{jobid}.err file yet does not contain any errors. That should indicate your server is up and running. Only until then can you attempt to connect to it. For example:

cat nb-log-{jobid}.err

[I 11:26:07.126 NotebookApp] Serving notebooks from local directory: /home/gaow
[I 11:26:07.126 NotebookApp] 0 active kernels 
[I 11:26:07.126 NotebookApp] The Jupyter Notebook is running at: http:// ...

4. After the server starts, you should see under the directory the log file of the job you just submitted:

cat nb-log-{jobid}.out

It should read like:

   Copy & paste the following to your local computer's terminal to tunnel 
to the remote server:

   ssh -N -L 9279:10.50.221.17:9279 <your username>@<the remote url> -v

   Then use a web browser on your local computer to open the following
address:
 
   localhost:9279

   You may be asked to provide a token (a string of 48 letters + digits). 
Please find it in the `nb-log-*.err` file under your current directory.

Just follow the prompts: run ssh -N -L ... command on your local computer, then open your faviorate web browser, eg, google-chrome, on your local computer and type in localhost: ... in the address bar.

5. The browser window will prompt you to type an access token to it. The information is in nb-log-{jobid}.err. In that file you should see a line that says:

Copy/paste this URL into your browser when you connect for the first time,
    to login with a token:
        http://<ip:port>/?token=xxxxxxxxxxxxxxxxxx

Copy & paste the token part xxxxx to the dialog box on your browser asking for the token, hit "enter", you should be directed to the Jupyter Lab server you created a moment ago.

6. To shut down the server, simply cancel the job

scancel {jobid}