Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 9.0-17
== General ==
- New Features
* Support for Rocky Linux and migrated to Rocky Linux clusters
* More information about migrating CentOS clusters to Rocky Linux is available in the Knowledge Base article
* https://kb.brightcomputing.com/knowledge-base/how-do-i-migrate-my-bright-centos-8-cluster-to-rocky-linux-8/
- Improvements
* Updated CUDA 11.4 packages to 11.4.2
- Fixed Issues
* An issue with setting up the Kubernetes Dashboard when using uppercase hostnames
* In some cases, an issue with compiling the cuda-driver
* An issue with the modules.sh causing an error on /bin/sh compatible shells like dash
== cmdaemon ==
- Improvements
* Improved client SSL certificate handling
* Fix a rare crash in monitoring entity/measurable linkage manager
* Prometheus cache optimizations
* Removed the option for using a null cipher for SSL traffic
* Fixed a possible race condition in taskrunner that can cause the monitoring samplers to stop working
- Fixed Issues
* Improved log information when cmdaemon cannot bind to a port
* An issue with the schedulers healthcheck for Grid Engine in the case when a queue is in a "Subordinated" state
* Updating the scheduler parameter in the slurmserver role does not restart the slurmctld service
* An issue with selecting the external network and gateway for an edge director when the BMC is also on the external network
* An issue with clearing the "Unknown error" state after a failed ping to a PDU
* An issue with Kubernetes label management when using mixed-case node names
* An issue with Jupyter authentication for users when PAM warnings are raised
* Potential retry loop that may continue regenerating Kubernetes certificates multiple times
* Rare crash in cloud storage manager
* In some cases cmdaemon may create a "no-version" WLM module file if /cm/shared is not (yet) mounted
* An issue with writing all interfaces' MACs in the dhcpd config file, which can result in a boot loop in the case of a bond interface
* Removed the Slurm prolog and epilog scripts from the exclude lists, i.e. they now can be rsynced from the software image also when using sync or update modes
== node-installer ==
- Fixed Issues
* In some cases, an issue with the standalone script on RHEL/Centos 8.4
* An issue with booting nodes with a bond interface with multiple VLANs
* An issue with getclientid script when using bonded Mellanox ConnectX4-Lx ethernet interfaces
== buildmaster ==
- Improvements
* Added hid-microsoft kernel module to default RHEL8/CentOS 8 image kernel modules
== cm-kubernetes ==
- Fixed Issues
* Ensure all Kubernetes images are prefixed with the full registry name
== cm-kubernetes-setup ==
- Fixed Issues
* prevent exception in cm-kubernetes-setup addon selection
== cmsh ==
- New Features
* Ability to list the available kernel versions for a software image
- Improvements
* Improved cmsh job filter command help