Last goodbye to the Skylake partition

The last skylake nodes will be decommission next Monday (November 10th) to make room for new GPU nodes.

Last July we started with the decommission of the old skylake nodes following on the Upgrade to Rocky Linux 9 carried out on our Tier-2 clusters. Just 16 nodes were kept online on a frozen state in the old system to help users transition to the new system. Now it is time to say the last goodbye to the skylakes. All Skylake nodes will be removed from the Hydra cluster next Monday (November 10th).

Users still relying on the skylake_mpi partition for their multi-node jobs can now target the zen5_mpi partition instead, which provides much faster CPUs and a much faster network interconnect. Alternatively, you can also remove any partition setting on your jobs and just let the job scheduler choose the best partition based on the resources requested in it.

Spoiler Alert: the space in the racks freed by the removal of the skylakes will be used by a new GPU partition for Hydra. More information coming soon.

Contact us at VUB-HPC Support if you have any comments or questions about these changes.