As you know, Docker containers cannot be run natively in HPC environments, but they can be easily converted and executed using Apptainer (Documentation). 

Currently an outdated apptainer installation is available without loading any module.

This installation of apptainer will be removed on 30th of November!

Please change your workflows to make use of more recent apptainer modules!

The November module updates and deprecations have been rolled out today. Please look out for deprecation warnings in your Slurm output.

Most notable new modules:

Common

aocc/5.0.0
aocl/ilp64/5.0.0
aocl/lp64/5.0.0
anaconda/2024.10
apptainer/1.3.5
cudnn/cu11x/9.5.1.17
cudnn/cu12x/9.5.1.17
micromamba/2.0.3
openjdk/8.u432-b06
openjdk/11.0.25+9
openjdk/17.0.13+11
openjdk/21.0.5+11

Scientific

gromacs/2024.4-ompi5.0-gcc13.2-mkl2023.2-cuda12.6
orca/6.0.1
siesta/5.2.0-ompi4.1-cf

The September module updates and deprecations have been rolled out yesterday. Please look out for deprecation warnings in your Slurm output.

Most notable new modules:

• cuda-compat modules for better compatibility using cuda toolskits > 12.2, see also https://collab.dvb.bayern/x/3PxdFw#NvidiaCUDAToolkit-CUDAToolkitinteroperability (cuda module needs to be loaded first)
• cuda-mps module for automatically starting and stopping CUDA Multi-Process-Service (highly recommended for code that cannot saturate A100 GPUs!), see also https://collab.dvb.bayern/x/m-xdFw#SubmittingGPUJobs-CUDAMulti-Process-Serverformaximumefficiency
• intel/2024.2.1 modules have been installed, intel/2024.1.0 is now deprecated, intel/2023.2.1 is still the default, see also here
• gcc/11.5.0 modules has been installed, gcc/11.4.0 is now deprecated
• orca/6.0.0 is available now, see also https://collab.dvb.bayern/x/yfxdFw

Since Friday, 30.08.,  a quota notification system is active.  Users will get an E-Mail message when their quota are exceeded. More information in our Knowledege Base: Quota regulations

On LiCCA a separate partition epyc-gpu-test has been created, and node licca047 with its 3 A100 GPUs moved to this partition. The TimeLimit in this partition is 6 hours, to give users the possibility to test with short job runs, while the bigger epyc-gpu is loaded with longer running jobs. If this is not really used, we will move GPUs (partially) back. All projects and users with GPU resources are automatically granted access to this  partition.

COMSOL 6.2 has been installed in the HPC cluster filesystem.

login node shell> ml load comsol/6.2
Loading Comsol 6.2

6.1 will still be the default for a few days, please give feedback if 6.2 installation still

has problems. 6.2 will switch to be default 1 or 2 weeks.

The HPC clusters LiCCA and ALCC are used with increasing
intensity and the load on the power distribution rises.

Stronger power cables have been laid for the power
infrastructure supporting the HPC clusters.

A shutdown of all compute nodes is scheduled for:

   Tuesday, May 14, 7:00

We will start to drain the queues on Saturday, May 11.

To the best of our knowledge the work on the
electrical system will be finished the same day,
so the clusters should be back on Wednesday, May 15.

We migrate the Slurm database instance (serving both the ALCC and the LiCCA cluster) to a different system starting today.

Slurm operation is planned to stay up during this time.

This should speed up Slurm operations after the migration, but

is also needed in preparation of the ALCC upgrade from Ubuntu 20.04 to 22.04, which will happen in the next weeks. 

we are proud to announce the availability of LiCCA, a compute resource focused on research, open for members of the University of Augsburg.

Access to LiCCA is possible after registration of your chair or working group for a HPC project. The complete application workflow is described in the HPC Knowledge Base, as well as the cluster hardware and setup.

Questions and problems, which are not solved within the HPC Knowledge Base, can be addressed to the Service-Desk at the Service- & Supportportal or by E-Mail.

Happy computing, the RZ HPC team