Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 9.2-13
== General ==
- Improvements
* Upgrade openssl to 3.0.9
* Updated cuda12.1 packages to 12.1 update 1
* Slurm: rely on MIG autodetection when configuring gres.conf
* Add mlnx-ofed23.04 package, version 23.04-1.1.3.0
* Allow cloud directors to not have public IPs for AWS Direct Connect cloudbursting
- Fixed Issues
* An issue with the AdditionalSubmitHosts advanced configuration flag
* A regression where CMDaemon restart triggers a Slurm restart
* Ensure job_gpu_* have correct values in the first few seconds of a job being started
* An issue where some events in cmsh were still displayed when running with --hide-events
* An issue where virtual IB interfaces were not excluded from monitoring
* An issue when initializing monitoring definitions from a JSON sampler
* An issue with with cm-kubernetes-setup logging creating too many files
* An issue when running cm-image shell
* An issue that prevented cm-jupyter-setup from running in multi-distro environments
* An issue fetching job info if UGE accounting rotation is configured
* An issue with the auto scaler ignoring queue priorities in case of multi-queue Slurm jobs
== CMDaemon ==
- Fixed Issues
* An issue with job_gpu_wasted metric calculation for multi-node jobs
* An issue where the NFSCheckerPort AdvancedConfig parameter was not being used
== Node Installer ==
- Fixed Issues
* An issue with the node-installer disk scripts being unable to assemble MD raids
* An issue with the bootif_detect and getclientid scripts not detecting ConnectX-3 cards while using the GRUB bootloader
== Head Node Installer ==
- Fixed Issues
* An issue with the selection of disks always using the first disk, regardless of user choice, if the custom disk editor was not opened.
== cmha-setup ==
- Fixed Issues
* An issue that prevented cloning headnodes containing btrfs filesystems
== jupyter ==
- Fixed Issues
* An issue where the internal jupyterhub database was not upgraded during package upgrades