Skip to content

Latest commit

 

History

History
30 lines (21 loc) · 1.1 KB

README_INTERNAL.md

File metadata and controls

30 lines (21 loc) · 1.1 KB
Raw

Blueprint for Slurm clusters with A3 compute nodes

Minimum Toolkit release

These blueprints are compatible with Cloud HPC Toolkit release 1.28.0 and above.

Notes

The A3 blueprints provision 2 sets of infrastructure

  • Base
    • VPCs and Filestore for /home/ directory
  • Cluster
    • Custom Slurm image based upon image including patched TCPx kernel
    • Slurm cluster itself

The blueprint must be used with appropriate settings for a3_reservation_name and a3_maintenance_interval. In particular, a3_maintenance_interval must be the empty string ("") in most ("General Fleet") projects and "PERIODIC" in projects that have maintenance preview APIs enabled. It is the responsibility of Google staff to identify the environment for the customer and recommend the appropriate value!

Please refer to the Google doc for further details regarding provisioning. It can be shared with customers.