Hi everyone, my name is Johannes Fee. I'm from the system team of NHR at FAU and I was
asked to present our JupyterHub server which had some changes over the past weeks. So let's go ahead.
So the first thing that changed is that it's now accessible on a different URL. So the one on top
of the slide is only for people with a legacy account. So tier 3 with a password. If you have
a new account which is provided by the portal for HPC, so the one linked here, then you have a nice
button once you've logged in which is screenshotted here on the left hand side and the right hand side
is where you put in your username and your password if you have one of the legacy accounts.
But we will see this in a moment in the live demo. Okay, then once you're logged in you will have
page something like this, so the landing page where you can pick between different kernel
options, different cluster configurations. So in the screenshot you can see some options from
TinyGPU and from TinyFAT if you don't have the need for GPU. And then there's a button which can
open up the additional menu options and one thing which is maybe interesting for the
teachers among us. So if you have a lecture which may or may not need Jupyter, you can ask us to
provide a magic prefix which provides additional configuration for example which loads the
Singularity container with your special software environment which is needed for the lecture.
So this is something you need to get in contact with us. The other ones should explain themselves,
they are basically the same option as you would use in your SLURM
job files. So the first one on top is loading additional modules. So if you need a special
compiler you can put it in there and it will be loaded once the Jupyter server starts up.
You can use a special reservation if there's one which you're allowed to use. The account one is
probably not interesting and you can change the directory. So normally your server will start up
in the home directory but you may have software you want to use in a different location.
Okay, so once you started your server and you did all your work on the Jupyter Hub
that you want to finish, please don't just close the browser tab because the server will
continue running in the background. So the GPU, the node, whatever you're allocated is still blocked
and this is wasting resources. Wasted resources are wasted money and we all know it's paid by taxes.
This is an easy way to save taxes. So once you've finished, go on the menu hub control panel and then
stop my server and this will kill the job in the background and free up the resources again.
Okay, so the list of kernels which can be used in Jupyter is more or less endless. So there's a
kernel for I think every programming language and for many programming languages there's more than
one kernel and obviously we can't maintain all of them publicly in all different configurations
which you can think of. So there's the option for everyone to just install the kernels they want to
use or you want to use and then it will be available as kernel option. So I put here example lines for
the condor installation point IPython kernel. So everything you need to do is add to your
environment. You may have IPython and it will show up in the Jupyter and for the app programming
language it's basically the same and then there's a link to all available kernels. So
feel free to try something out if you need help contact us.
Then we have a lot of additional documentation on our work pages which is linked here and obviously
there's tons of documentation by the community which might not always work one-to-one on our
system because we have some special configurations here and there but most of it can be used or
with small adaptation used. So yeah try to use all the documentation and if there's a problem
feel free to contact us and we will try to help you. Okay so let's move on with the live
demonstration because this is where the fun starts. So as you can see we have my browser tab here
as you can see we have my browser tab here and so if you just go to the web page linked you will be
welcomed by this page. Please when you log in for the first time have a look at these
the text on top it contains the terms of use and some further information you should
know of before you use the system. Once you have acknowledged this and read through it you can
put in here your username and your password like I did login and then start my server and we will
be at this page you already saw on the screenshot and now we have the possibility to use whatever
resources we think is useful for us. So for the sake of demonstration I will use the local server
Presenters
Zugänglich über
Offener Zugang
Dauer
00:31:25 Min
Aufnahmedatum
2023-07-12
Hochgeladen am
2023-07-12 09:26:03
Sprache
en-US
Topic: Using the Jupyterhub service at NHR@FAU
Speaker: Dr. Johannes Veh, NHR@FAU
Abstract: Jupyter notebooks are growing in popularity for python, Julia and R development, thanks to interactive code execution. Jupyterhub brings the power of notebooks to groups of users. The Jupyterhub service at NHR@FAU has seen several changes over the past weeks. We will bring you up to speed, from password-less login to custom jobs on your favorite cluster.
Material from past events is available at: https://hpc.fau.de/teaching/hpc-cafe/