1 PC² JupyterHub
- 1.1 Access
- 1.2 Quick Start
- 1.3 Server Options
  - 1.3.1 Presets
  - 1.3.2 Simple
  - 1.3.3 Advanced (Slurm)
- 1.4 Loading additional Jupyter kernels
- 1.5 Apptainer (Singularity) Container
- 1.6 Remote Desktop (Graphical Environment via Xpra)
- 1.7 How-To
  - 1.7.1 Creating presets
    - 1.7.1.1 Spawner
    - 1.7.1.2 Preset scopes
    - 1.7.1.3 Default URL
    - 1.7.1.4 Notebook directory
    - 1.7.1.5 Apptainer container
    - 1.7.1.6 Environment variables
    - 1.7.1.7 Modules
    - 1.7.1.8 Slurm job flags
  - 1.7.2 Create custom IPython kernel inside custom conda environment
  - 1.7.3 Create my own Apptainer/Singularity container
    - 1.7.3.1 Container package requirements
    - 1.7.3.2 Example Apptainer/Singularity recipe
      - 1.7.3.2.1 Base recipe
        1.7.3.2.1.1 Install custom Python kernel inside the container (python 3.12)
        1.7.3.2.1.2 Install Lmod with the JupyterLab-Lmod extension
        1.7.3.2.1.3 Make Slurm Tools inside my container available
    - 1.7.3.3 Container Location
  - 1.7.4 Access remote JupyterHub server with the local Visual Studio Code instance
  - 1.7.5 View Slurm job logs
  - 1.7.6 PC² Support
- 1.8 Troubleshooting
  - 1.8.1 JupyterLab
    - 1.8.1.1 “A Jupyter server is running.” message

PC² JupyterHub

The JupyterHub service is available for Noctua 1 and Noctua 2.

Access

The JupyterHub can be reached at the following address:

Noctua 2: https://jh.pc2.uni-paderborn.de

Noctua 1: https://jh.noctua1.pc2.uni-paderborn.de

The JupyterHub can be accessed via VPN or on-site at the University of Paderborn.

Quick Start

Spawn host/resources	Start

Spawn host/resources	Start
Jupyter Session on Noctua 2	Quick Start
Jupyter Session on Noctua 1	Quick Start
Jupyter Notebook on Noctua 2 (Inside Slurm job, 1h runtime, normal partition)	Quick Start
Jupyter Notebook on Noctua 2 (Inside Slurm job, 1h runtime, gpu partition)	Quick Start
Jupyter Notebook on Noctua 1 (Inside Slurm job, 1h runtime, normal partition)	Quick Start
Jupyter Notebook on Noctua 1 (Inside Slurm job, 1h runtime, gpu partition - 1x A40)	Quick Start

Server Options

Presets

You have the possibility of creating a preset with predefined start options for yourself or your project group.

Note: The preset functionality is currently only available on Noctua 2.

Click here to list your presets: https://jh.pc2.uni-paderborn.de/services/presets/

Simple

Preset enviroments with predefined values how to start the Jupyter Notebook.

Default and self-created apptainer containers can be used.

Advanced (Slurm)

An advanced view with setting options how a Slurm job should be started on a HPC cluster.

Loading additional Jupyter kernels

You can load additional Jupyter kernel using Lmod (module). Following kernel are currently available:

If you need new kernel versions or even other programming languages then you are welcome to contact pc2-support!

Apptainer (Singularity) Container

In JupyterHub it is possible to launch Jupyter Notebook instances inside a Singularity container. This has the advantage of being able to use your own built environment. When starting a container, any directories can be mounted inside the container environment.

We provide a set of default Singularity containers:

Container name	Kernels available	Installed software

Container name	Kernels available	Installed software
jupyter_scientific_python	Python	includes popular packages from the scientific Python ecosystem (more here: Selecting an Image — Docker Stacks documentation)
jupyter_datascience	Julia, Python, R	All from “jupyter_scientifc_python” rpy2 package The Julia compiler and base environment IJulia to support Julia code in Jupyter notebooks HDF5, Gadfly, RDatasets packages

To learn more about Singularity, see here: Singularity-Introduction

If you want to build your own Singularity container for JupyterHub, see here: JupyterHub | Create my own Singularity container

Remote Desktop (Graphical Environment via Xpra)

To create a remote desktop environment, you can click on "Desktop Environment" in the JupyterLab interface:

When you click on the tile ‘Xpra Desktop’, a remote desktop environment is set up in the background. Graphical applications (e.g. loaded via modules) can be started from the started graphical terminal.

How-To

Creating presets

Note: The preset functionality is currently only available on Noctua 2.

To save time when configuring your Jupyter environenment you have the possibilty to create preset environments for yourself or your compute time group(s).

Created presets can be selected when starting a new Jupyter instance:

Create presets here: https://jh.pc2.uni-paderborn.de/services/presets/ (or JupyterHub home → services → presets)

Spawner

Local spawner (on JupyterHub)
- Spawning the Jupyter notebook environment on the JupyterHub host. Slurm job flags not needed.
  - Slurm tools, Modules, Remote desktop environment are available.
Noctua 2 (via Slurm)
- Spawning the Jupyter environment inside a Slurm job (on a compute/gpu/fpga node) on Noctua 2. Note: You need to specifiy Slurm job flags.

Preset scopes

Select who can use your preset. You or one of your compute time projects.

Default URL

The URL to which JupyterHub redirects when the server is started.

Example:

/lab -> Spawning JupyterLab environment

/xprahtml5 -> Spawning Remote desktop environment

Notebook directory

The working directory. Used for JupyterLab, the remote desktop environment and the classic Jupyter view.

Apptainer container

Your self-built Apptainer/Singularity container. Have a look here for creating your own container: JupyterHub | Create my own Apptainer/Singularity container

Environment variables

Extra environment variables.

Format:

MY_ENV_VAR=”Hello World”
FOO=BAR

Modules

Extra Lmod modules to load on start time. All system modules and Jupyter specific kernels are available.

Slurm job flags

Slurm job flags in Slurm batch format. Example:

#SBATCH --partition=normal
#SBATCH --time=01:00:00

Create custom IPython kernel inside custom conda environment

Create a conda environment as described here:
1. Python & Python Package Management
conda activate <your_conda_env>
conda install ipykernel
1. Or: python3 -m pip install ipykernel
python3 -m ipykernel install --user --name <KERNELNAME> --display-name "<DISPLAY NAME>"
1. Make sure, that python3 is called from the environment

Create my own Apptainer/Singularity container

Container package requirements

python >= 3.10
- jupyterhub
- optional, but useful: jupyterlab

Example Apptainer/Singularity recipe

Build containers: Apptainer

Base recipe

Bootstrap: docker
From: debian

%post
apt -y update
export DEBIAN_FRONTEND=noninteractive
apt -y install zsh locales
localedef -i en_US -c -f UTF-8 -A /usr/share/locale/locale.alias en_US.UTF-8

python3 -m pip install jupyterhub

Install custom Python kernel inside the container (python 3.12)

mkdir /opt/python3.12
cd /opt/python3.12

apt -y install build-essential libssl-dev zlib1g-dev libbz2-dev libreadline-dev libsqlite3-dev wget curl llvm libncurses5-dev libncursesw5-dev xz-utils tk-dev libffi-dev liblzma-dev python3-openssl git

wget https://www.python.org/ftp/python/3.12.0/Python-3.12.0.tgz
tar -xf Python-3.12.0.tgz
rm Python-3.12.0.tgz
cd Python-3.12.0/
./configure --enable-optimizations
make -j 8
make altinstall

python3.12 --version
python3.12 -m pip install --upgrade pip
python3.12 -m pip install ipykernel

# finally installing ipython kernel
python3.12 -m ipykernel install --sys-prefix  --name <UNIQUE_KERNEL_NAME> --display-name "<KERNEL DISPLAY NAME>"

Install Lmod with the JupyterLab-Lmod extension

apt -y install lua5.3 lua-posix

mkdir -p /usr/lib64/lua/5.3
cp /usr/lib/x86_64-linux-gnu/liblua5.3-posix.so.1 /lib64/lua/5.3/
mv /lib64/lua/5.3/liblua5.3-posix.so.1 /lib64/lua/5.3/posix.so

Make Slurm Tools inside my container available

groupadd --gid 351 munge
groupadd --gid 567 slurm
useradd -d /var/run/munge -M --gid 351 --uid 994 --shell /sbin/nologin munge
useradd -d /opt/software/slurm -M --gid 567 --uid 567 --shell /bin/false slurm

Container Location

All containers with type .sif will be automatically detected in $HOME/.jupyter/pc2-jupyterhub/

Your new built container can only be placed in your $HOME directory: $HOME/.jupyter/pc2-jupyterhub/

Alternatively you can create a link from your $PC2PFS to your $HOME directory:

ln -s /scratch/hpc-prf-project/jupyter_container.sif $HOME/.jupyter/pc2-jupyterhub/

Access remote JupyterHub server with the local Visual Studio Code instance

You need following extensions for Visual Studio Code:

Create an acccess token in the JupyterHub web interface:

Bildschirmfoto 2024-07-31 um 11.04.32.png

Follow following instructions described here: https://marketplace.visualstudio.com/items?itemName=ms-toolsai.jupyter-hub
1. You need to start a Jupyter session using our JupyterHub web interface. After successful start, you can copy the URL starting with https://jh.pc2.uni-paderborn.de/user/.../...

View Slurm job logs

If the path of the Slurm Job output has not been changed explicity, it can be found here by default:

Noctua 1: $HOME/.jupyter/last_jh_noctua1.log

Noctua 2: $HOME/.last_jh_noctua2.log

PC² Support

If you have any other problems that won’t be solved, please contact the pc2-support@uni-paderborn.de

Troubleshooting

JupyterLab

“A Jupyter server is running.” message

This message appears because user settings managed by JupyterLab do not match the new JupyterLab version.
- Try deleting ~/.jupyter/ in your $HOME directory as follows:
  - rm -r ~/.jupyter/
- If you want to keep your custom user settings, write an email to pc2-support@uni-paderborn.de

PC2-Documentation

JupyterHub