Base Command Manager / Bright Cluster Manager Release Notes

Release notes for Bright 9.1-3

== General ==

- Fixed Issues

* AdvancedConfig = { "SSLServerMethod=TLS 1.2" } no longer implies 1.2 + 1.3, it is recommended to use AdvancedConfig = { "SSLServerMethod=TLS 1.3" } instead
* To achieve the old behavior of 1.2 + 1.3, AdvancedConfig can be set to { "TLS 1.1=0", "TLS 1.2=1", "TLS 1.3=1" }

- New Features

* Added Slurm 20.11

- Improvements

* Updated cuda-driver to version 460.32.03
* Major Slurm versions can now be upgraded without disabling the old one (see the documentation)

== cmdaemon ==

- New Features

* Ability to work with the default cgroup layout when collecting WLM job metrics, without the use of JoinController systemd parameter

- Fixed Issues

* Possible deadlock in cmd stop when power manager is in heavy use at the time
* An issue with job metric collection for the special case when the job process forks a new process, but without WLM environment variables
* An issue with starting nightly provisioning updates
* Do not add Slurm gres Cores setting to gres.conf when it is empty
* An issue with software image remove with all revisions, which can sometimes crash cmd
* In some cases, prejob healthcheck tries to run on the node before the monitoring controller on the node is started
* An issue with createramdisk on fully updated CentOS 7 head nodes with FIPS enabled
* Do not convert Slurm gres count if it is not divisible by 1024
* Prometheus date with timestamp sometimes gets added out of order

== Bright View ==

- Fixed Issues

* Software image creation time shows incorrect date

== bright-installer ==

- Fixed Issues

* bright-installer (addon): Fixed an issue with detecting the Bright ISOs versions

== cm-kubernetes-setup ==

- Improvements

* Fixed a potential crash in the wizard regarding auto-detection of available network CIDR ranges

== cm-scale ==

- Improvements

* Allow for keeping (not terminating) specified cloud nodes when they are unused

- Fixed Issues

* Don't undrain a node that was drained by WLM due to prejob healthcheck (prolog) failure

== cm-setup ==

- New Features

* cm-edge-setup: Allow for different categories to be specified for edge director and edge nodes

== cm-wlm-setup ==

- Improvements

* Support for Slurm 20.11

== cmsh ==

- Improvements

* Added cmsh flag to clear all history
* List only the used regions and types with cmsh cloud prices command

- Fixed Issues

* An issue with setting the time zone with cmsh

== wlm-setup ==

- Fixed Issues

* An issue with usernodelogin=onlywhenjob when a Slurm user runs salloc