Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 9.0-5
== General ==
- New Features
* Added support for RHEL 8.1 AARCH64
- Improvements
* Updated cuda-driver to version 440.64.00
* Updated PGI compiler suite to version 19.10
* Added mlnx-ofed50 packages
- Fixed Issues
* Updated openmpi3 packages to version 3.1.6, to solve glibc issue on RHEL 8 and derivatives.
== cmdaemon ==
- New Features
* Improved LSF dynamic hosts handling
- Improvements
* Fixed rare deadlock in status manager during cmdeamon stop
- Fixed Issues
* An issue with executing state flapping actions
* An issue with making the previous active of a failover group to be active again when all nodes were down at the same time
* Limit the images for which the edge director can perform operations to the images listed in the provisionig role
* Use the partition relay host also for the compute nodes
* Allow HA take over with pending provisioning requests if the target node is not UP
== node-installer ==
- New Features
* Ability to create (copy) category- and node-specific configuration files
== Bright View ==
- Improvements
* Show the account name of the logged in user
* Improved default WAR report
- Fixed Issues
* Unresponsive settings form for JupyterHub role
* In some cases, issue with WLM wizard unable to complete
* Issue when editing cluster admin email address
* SoftwareImage creation time input shows raw timestamp values
* Create a new monitoring tree when an invalid cache ID is passed
* Some jobs metrics can show 'undefined' in the chart's label name
* Mouse zoom in Rack view not working with Firefox browser
* Job queue stats shows duplicates values
* An issue with showing/editing node groups
* Rack-view: Improve labels for unassigned devices racks
* Missing configuration input fields for Kubernetes AppGroups
* Fixes for WAR charts legends layouts and performance
* Hovering of chart's data points not working in Firefox browser
== cm-cluster-extension ==
- Fixed Issues
* An issue which can cause cm-cluster-extension to sometimes hang
== cm-scale ==
- Fixed Issues
* cm-scale doesn't produce an error for invalid node ranges
== cmsh ==
- New Features
* New --wait flag for compute node makeactive
- Improvements
* New --wait option for creating new revisions
== head node installer client ==
- Fixed Issues
* Custom network gets renamed to the default value when navigating through installer steps
== ml ==
- New Features
* Introduced machine learning package cm-cub-cuda10.2
== monitoring ==
- New Features
* Allow price to be specified for a prometheus query so that the cost can be calculated
* New average job metrics to WLM cluster entity
- Improvements
* Allow monitoring resources to be specified for fsmounts and fsexports
* New query to calculate the wasted memory usage for users' running jobs
- Fixed Issues
* Support for promql on openstack data
== pbspro ==
- New Features
* Upgrade pbspro19 package to 19.2.6
== pythoncm ==
- Improvements
* Prevent duplicate connections
- Fixed Issues
* An issue with using images with revisions
== user portal ==
- Fixed Issues
* Some jobs metrics can show 'undefined' in the chart's label name
* Hovering of chart's data points not working in Firefox browser