This directory stores codes and files that enable to tunnel jupyter from compute nodes to a local computer. Therefore, we use a slurm job file, that can be started using an alias.

The procedure is inspired by https://nero-docs.stanford.edu/jupyter-slurm.html .

For the use of this framework, you will need to rund commands on the SERVER, as well as on your LOCALcomputer, which are highlited in the headings to avoid confusions.

Installation - SERVER + LOCAL

Clone the repository both, on server and on local computer, using

git clone https://github.com/burki-lab/juppmaxyter_tunnel.git

Ensure to have installed (or loaded) JupyterLab on the server.

Run Jupyter on server

Follow the next steps to start Jupyterlab on computation nodes of SLURM server.

Customize the config-file - SERVER + LOCAL

In config.yml you should store your specific configurations on the server, as well as on your local computer. Set up the file on one of the computation units (e.g. server) and copy its content to the other one (hence, the local computer).

Start the job on the server - SERVER

After configuration, run src/start_jupy_job.sh on the server. This will submit a job using your configurations to start JupyterLab on a computation node.

Wait until job is started ... - SERVER

Using jobinfo -u $USER you can check, if the job is already started.

Make the ssh-tunnel - LOCAL

Now, there should be a running JupyterLab job. We can tunnel it using src/copy_logs_via_ssh.sh on the local computer. If you are using a password secure server, you will twice be asked for a password. Before the second prompt you should get a link that you could copy like

Copy to browser: <LINK>

Open JupyterLab in browser - LOCAL

Copy the link from before to your favorite browser and enjoy the new Unyverse.

Terminate JupyterLab - SERVER

To save compuation hours always remember to cancel the Jobs and delete the logs. This can be done using src/cancel_jupyter_jobs.sh.

Practicality

To increase the usefullness of juppmaxyter_tunnel, one could add aliases to the shell source file (e.g. by using vim ~/.bashrc and adding new the alias lines) for fast and simple access:

• SERVER:

• • for changing of configs alias 'jupchange=vim <ABSPATH_JUPPMAXYTER_TUNNEL_SERVER>/config.yml'
• • for starting jobs alias 'jupy=<ABSPATH_JUPPMAXYTER_TUNNEL_SERVER>/src/start_jupy_job.sh'
• • for cancelling all jobs alias 'jupterm=<ABSPATH_JUPPMAXYTER_TUNNEL_SERVER>/src/cancel_jupyter_jobs.sh'

• LOCAL:

• • for changing of configs alias 'jupchange=vim <ABSPATH_JUPPMAXYTER_TUNNEL_LOCAL>/config.yml'
• • for starting the tunnel alias 'juptunnel=<ABSPATH_JUPPMAXYTER_TUNNEL_LOCAL>/src/copy_logs_via_ssh.sh'