v2.8
What's Changed
Improved GPU configuration/support
- Support automatic GRES configuration for NVIDIA GPUs by @sjpb in #820
- Add option to install nvidia-fabricmanager by @claudia-lola in #836
- Add support for GRES to ondemand apps by @sjpb in #837
- Adds bandwidth.yml playbook for NVIDIA nvbandwidth by @claudia-lola in #834
- Make eessi configure gpu node automatically by @claudia-lola in #841
- Bump CUDA to 13.0.2 and NVIDIA driver to 580.105.08 by @priteau in #823
Slurm configuration
- Bump OpenHPC role to v1.4.0 by @sjpb in #818. Adds:
- Enable use of custom Slurm builds by @sjpb in stackhpc/ansible-role-openhpc#163
- CI: Switch to latest rockylinux/rockylinux images by @priteau in stackhpc/ansible-role-openhpc#198
- Add support for mpi.conf templating by @bertiethorpe in stackhpc/ansible-role-openhpc#201
- Bump OpenHPC role to v1.4.1 by @bertiethorpe in #822 - fixes mpi.conf templating
New/improved features
- Add support for InfiniBand interfaces to NHC by @sjpb in #821
- Add tool to set image properties by @sjpb in #829
Docs and other
- Improve pulp docs by @sjpb in #819
- Fix gpg check for cernvmfs installs by @bertiethorpe in #816
- Remove ansible-lint warnings by @bertiethorpe in #817
- Replace whitespace in NHC mount checks by @sjpb in #824
- Allow fixed ip lists to be longer than nodes list by @sjpb in #830
- Add retries to CI tofu apply by @bertiethorpe in #833
- Don't install hpl source during extra builds by @sjpb in #828
- Add docs for eessi by @claudia-lola in #827
- Fix ansible-ssh changes due to linting by @sjpb in #838
- Use Ark repofiles for additional repos by @bertiethorpe in #832
- Set image properties for CI image build and sync by @bertiethorpe in #839
- Run trivy scans on main, to help reporting by @JohnGarbutt in #842
- Describe buildenv in EESSI docs by @claudia-lola in #845
Full Changelog: v2.7...v2.8
Images
Two new images are available:
- RL8: openhpc-RL8-251119-1202-332ac921
- RL9: openhpc-RL9-251119-1202-332ac921