CHEESE Advanced Training on HPC for Computational Seismology

https://cheese-coe.eu/events/cheese-advanced-training-hpc-computational-seismology

The Salvus tutorial happens October 21rd from 9:00 to 12:00 Central European Time (CET).

Opening a Jupyter Notebook for the Salvus tutorial

A prepared installation has been set up on the training cluster of the HLRS so the only thing that remains to be done is to launch it and connect it to your local browser.

Your SSH connection should already be set up from all the other tutorials so this should be quick.

Linux/OSX/WSL

Open a shell, copy the following line (replace scaXXXXX with your actual user name!) and hit enter.

Copy
USER=scaXXXXX; ssh -L 33${USER:(-3)}:localhost:33${USER:(-3)} -t ${USER}@training.hlrs.de "bash /shared/cheese/salvus/config/launch_salvus.sh 33${USER:(-3)}"

This will take around 30 seconds and at the end it will present you with a link like https://localhost:XXXXX/?token=.... Open this link in your browser and you are all set.

Windows

This is a bit more complicated, if possible we'd recommend to use the WSL on Windows and do what is written above.

But PuTTY can also be made to work (most things should already be set up from the previous tutorials):

1. Hostname + port

2. Username

3. Remote command

You can either enter this here (it will then be run automatically upon connecting) or just run it once the remote connection opened:

Replace the 33XXX as written above.

4. Key

This should already be set up from the previous tutorials

5. Tunnels

Add 33XXX as the source port and localhost:33XXX as the destination and then press Add.

Replace the 33XXX as written above.

6. Save and Open Connection

Save the session (to not have to re-enter everything if something is wrong) and click "Open".

This will take around 30 seconds and at the end it will present you with a link like https://localhost:XXXXX/?token=.... Open this link in your browser and you are all set.

What is happening here?

It opens an SSH connection and directly executes the launch_salvus.sh script. Additionally it forwards a local port to the same port number on the remote machine.

Upon execution the script prepares your environment on the training cluster for Salvus, launches an interactive job for the duration of the tutorial, tunnels the same port from the compute to the login node, and finally launches a jupyter notebook server on that same port.

All the tunneling really only serves the purpose to be able to access the jupyter notebook server from your local browser.

tmux

The script internally runs everything in tmux - thus if your connection drops for some reason, just rerun the script and will detect that tmux runs and only keep the tunnel open. Thus you should be able to reconnect to the jupyter notebook server with your browser.