Overview

Users may login via ssh:

ssh <jmu_account>@julia2.hpc.uni-wuerzburg.de
  • On Linux systems you can use ssh in your terminal directly.
  • Apple users have ssh installed and also can use ssh directly in the terminal.
  • Windows users (Windows 10 and above) can use ssh directly via powershell or by using a ssh client like putty or MobaXterm, which also can be purchased in the webshop if you need a full license.

Hint:

Your <jmu_account> is usually in the form sXXXXXX, where X are numbers e.g. s123456 for students or abcXXde, e.g. der42py for employees. Do not try your email address.

Login Node

The login node should be used for:

  • Setup your environment
  • Transfer your data
  • Compile your code
  • Submit your SLURM jobs

It should not be used for:

  • Running jobs directly on that node
  • Testing if your stuff works

Limitations:

The login node only has 64 GB of RAM and 16 cores. Users are limited to 16 GB of memory and 8 cores for usage.

Setup your environment

You can download your code to your home directory and compile your code here. See also the limitations above!

Internal SSH keys

In order to be able to login to compute nodes for debugging, please create a pair of ssh-keys. You can do that on the login node via:

ssh-keygen -t rsa
cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys

For the simple case, you can just answer all questions by hitting the ENTER key when asked by ssh-keygen. The cat command will add the internal ssh key to your authorized_keys if it already exists or create a new one.

Example: python environment

Create a new virtual python environment, where you can install the packages you require.
Ignore all the comments: // This is a comment 

mkdir project                                // Create a project directory
cd project                                   // Switch to project directory
python3 -m venv .                            // Init a new python environment
source bin/activate                          // Activate the new python environment
pip3 install torch                           // Install a python package, e.g. pytorch

Example: Install gromacs

Follow the steps described on https://manual.gromacs.org/current/install-guide/index.html

mkdir $HOME/bin                     // Create a bin folder, where we install the gromacs binary later
export PATH=$HOME/bin:/usr/local/cuda/bin:$PATH  // Add $HOME/bin to your PATH variable and also include the CUDA compiler if needed
tar xfz gromacs-2024.4.tar.gz       // Extract the gromacs archive after downloading
cd gromacs-2024.4                   // Switch to the gromacs directory
mkdir build                         // Create build folder, needed for cmake
cd build                            // Switch to the build folder
cmake .. -DGMX_BUILD_OWN_FFTW=ON -DREGRESSIONTEST_DOWNLOAD=ON -DGMX_MPI=on -DGMX_GPU=CUDA -DCMAKE_INSTALL_PREFIX=$HOME/bin                // Run cmake. Also enable MPI or CUDA, if needed
make                                // Compile gromacs .. Go and bake a cake .. this takes a while
make check                          // Run the checks
make install                        // Install GROMACS in your $HOME/bin folder
source $HOME/bin/GMXRC              // Init gromacs environment

To add the gromacs path and also the CUDA compiler permanently to your PATH variable run

echo 'export PATH=$HOME/bin:/usr/local/cuda/bin:$PATH'  | tee -a $HOME/.bashrc

Example: Singularity/Apptainer

Check the Running container page for examples how to build and run containers on Julia 2.

Diskspace

Home

137 TB in Total on NVMEs

/home/<user> has a default quota of 150GB

Use for: code and environments

Data

1.8 PB in Total on harddrives

/data/<primary_group>/<user> has no quota set yet.

Use for: Compute data

Test your stuff

Please use the test partition (#SBATCH -p test) to see if your stuff is running before submitting your jobs to the cluster.

The jntest01 node has 2 GPUs and 64 cores. Accessing the node can be done by an interactive job.

Be aware: Running jobs on the test partition are currently limited to one hour of usage.

See the Basic Jobs page for example job files.