Spartan has been upgraded to RedHat Enterprise Linux 9, Slurm 23.02.5 and Spectrum Scale 220.127.116.11
There are many changes to the system which you should read and get accustomed with before submitting jobs.
Please be patient with us when the system comes back online. The number, size and complexity of changes we made means that there will probably be things that don't quite work, despite our extensive preparation and testing. Please submit a ticket if things aren't quite working well, and describe your issue in as much detail as you can (including modules being loaded, job number and error message seen).
New software system
The old software systems in RedHat 7 are no longer available. The new software system is based on hierarchies, where you can only see the software in the toolchain you have loaded. The new software system is case sensitive. See Modules for details.
On that page, you will see suggested module load statements for different workflows.
fosscuda has been removed. To use GPU software, load a CUDA version, and look for modules with CUDA in the name.
Singularity has been removed. The Singularity project forked into 2 - SingularityCE and Apptainer. We have chosen to install Apptainer, which can be used in identical fashion to Singularity. Load the
Apptainermodule if you wish to use containers on Spartan.
In anticipation for new hardware to arrive this year, which will have different CPUs than our current hardware, we have renamed
cascade(the current CPUs are Cascade Lake CPUs). Please see Specifications for details.
Removal of FastX
We have removed the old FastX system for remote desktops. Remote Desktops are now available through Open OnDemand, including a GPU enabled desktop option. See Open OnDemand for details.
Recommendations for returning users:
Delete your R libraries and Python environments, and recreate.
R libraries are by default stored in $HOME/R.
Python environments are stored in $HOME/.local (for pip install --user), and $HOME/venvs for virtual envs. Delete these directories, and recreate the environments you require.
pip install --userto install Python modules. We highly recommend you move to using virtualenvs and/or Conda environments for Python module installation. This is much neater and allows you to separate Python modules for different tasks. See Python for details.