Base Command Manager / Bright Cluster Manager Release Notes

Release notes for Bright 6.1-14

== cmdaemon ==

* Improved: SLURM job info parsing: use ReqNodeList to get nodes required by the job. If ReqNodeList is not defined, then NodeList will be used.
* Improved: a MIC device will not be added to the slurm's gres.conf if for the MIC's SlurmClient role the Gres property is set to 'no'
* Fixed: in some cases, uplink settings being lost when additional uplink ports of Ethernet switches are configured using cmgui
* Fixed: possible corruption of fspart associations and cmdaemon crash when removing an image and committing a node/category quickly one after the other
* Fixed: an issue with MPSS being reconfigured while it is starting, which can lead to an incomplete micX.filelist and prevent MICs from booting
* Fixed: MIC nodes being counted as regular nodes while reporting billable nodes
* Fixed: rogueprocess healthcheck reporting UNKNOWN if a process has terminated while the check was running

== node-installer ==

* Fixed: a problem that prevented cloning a head node that uses a GPT based disk setup.

== manuals ==

* Added: the check command for health checks on demand

== buildmaster ==

* Added: yum-(plugin-)priorities package to the default package installation list for the head node and the software image for RHEL-based distributions

== cluster-tools ==

* Fixed: newly added MICs remaining in DOWN state

== cm-config-yum ==

* Set the priority for Bright Computing RPM repositories to 11

== cm-config-zypper ==

* Set the priority for Bright Computing RPM repositories to 11

== cmgui ==

* Fixed: start, stop, and restart of OS services not working properly in some cases
* Fixed: slots property not being displayed for Slurm client role
* Fixed: disksetup parsing not working if raid priority is specified

== cuda-driver ==

* Update of driver and libs, from 319.49 to 319.60.

== intel-mic ==

* Added: intel-mic-ofed package dependency: libibverbs-devel
* Fixed: loading ofed drivers for Mellanox HCA