Date:
02/12/2024
After nearly a decade of service to researchers at the University of Minnesota, the Mesabi computing cluster will be retired on June 5, 2024 and MSI’s clusters will be reconfigured during Summer 2024. The Mangi nodes, which have been attached to Mesabi, will remain in service and be attached to Agate under the new arrangement. Agate will also be expanded with newly purchased nodes in the late summer of 2024.
Impacts for MSI users
SLURM Partitions:
SLURM Partitions retiring on June 5:
- large
- small
- ram256g
- ram1t
- k40
SLURM Partitions changing on June 5:
- max -> moving to AMD CPU nodes on Agate
Partitions to be deprecated (still recognized by SLURM, but not listed in documentation):
- amdsmall
- amdlarge
- amd512
- amd2tb
- v100
Software:
- MSI will rebuild software targeting Mesabi as needed
- Users with their own software targeting Mesabi will be notified about recompiling their code
- Related: MSI systems will be upgraded from Centos7 to Rocky8 during the May maintenance (Centos7 is reaching end-of-life)
- Two solutions for modules that can’t run on Rocky8: rebuild in the new environment, or have a compatibility layer via an apptainer image
- The above solutions should also be used for user-built software targeting Centos 7