Base Command Manager / Bright Cluster Manager Release Notes

Release notes for Bright 7.1-24

== cmdaemon ==

* Added: adv. config. option AddUserDefaultGroupID to set the default primary group when adding new users
* Added: cm-list-managed-files script to list all files updated by CMDaemon
* Improved: ensure Ceph health check displays a clear message when an admin socket does not exist
* Improved: handling of number ranges for GPU settings
* Improved: display the number of pay-per-use nodes in cmsh main licenseinfo
* Fixed: node features were being added to the GresTypes list in slurm.conf
* Fixed: an issue with SGE/OGS nodes no longer drained when moved to another queue
* Fixed: an issue with applying cpu scaling governor setting
* Fixed: pxelabel is not written to pxelinux.cfg files
* Fixed: GE integration: additional submit hosts are not added properly when short hostnames are used
* Fixed: /etc/hosts does not contain the correct master IP for nodes with defined management network
* Fixed: an issue which in rare cases can lead to compute node cmdaemon ending up in an infinite loop of processUpdateRemoved resulting in an increased number of threads and CPU load
* Fixed: CMDaemon adds CMD variables to the environment variables of the services that are started by it
* Fixed: pexec -s is broken
* Fixed: cmsh rshell connects to head node if current node is down
* Fixed: CMDaemon doesn't start second HA interface if first is already running
* Fixed: cluster.find(name, baseType) not working for numeric names in import1xml example (additional fix)
* Fixed: import1xml can't import node because it doesn't find the provisioning interface
* Fixed: gpuhealth healthcheck now reports unknown if it can not determine a the result of an known nvidia-healthmon sub-test.
* Fixed: an issue with adding user groups with names longer that 15 characters
* Fixed: ngpus and gpus complex value set for SGE and UGE.

== node-installer ==

* Added: support for OEL when creating ramdisks
* Fixed: in some cases, node-installer syslog entries containing 'localhost' or '(none)' as the hostname. Now changed to logging with IP address
* Fixed: previously kernel parameters were not set in the node-installer initrd for sles11, rhel6 and derivatives. Now for sles11, rhel6 and derivatives kernel parameters, which are set in sysctl config files of the image, are set in the node-installer initrd. (Kernel parameters are set by default in the initrd for sles12, rhel7 and derivatives.)
* Fixed: Infiniband node_desc isn't set when the node installer is running

== admin-manual ==

http://support.brightcomputing.com/manuals/7.1/admin-manual.pdf

* Updated: Torque 5+ and Maui are currently incompatible

== cluster-tools ==

* Added: config files for cm-create-image to create OEL7 software images.

== cm-config-man ==

* Fixed: manpath warning when setting MANPATH variable

== cm-curl ==

* Updated to 7.48.0

== cm-scale-cluster ==

* Fixed: calculation of nodes that can be terminated
* Fixed: terminating nodes number calculation when some nodes are busy
* Fixed: crash when ASSIGN_CATEGORY parameter is applied to template nodes that does not exist yet

== cmgui ==

* Improved: display the number of pay-per-use nodes in license info
* Fixed: an issue with using cmgui with Firefox 46.0
* Fixed: Setting image revision for cluster default image doesn't work
* Fixed: In Rackview, heathchecks which return PASS are displayed in blue instead of green

== cuda-driver ==

* Updated: to version 352.93

== intel-compilers-2016 ==

* Updated: compiler suite to version 2016 update 3.

== intel-mic ==

* Updated: to MPSS version 3.7.1.

== intel-tbb ==

* Updated: to version 44_20160526oss.

== intel-truescale-ofed ==

* Updated: to version 7.4.1.0.24.

== intel-wired-ethernet-drivers ==

* Updated: igb driver to 5.3.5.3, ixgbe driver to 4.4.6, ixgbevf driver to 3.2.2, i40e driver to 1.5.18, i40evf driver to 1.5.14, fm10k driver to 0.19.6.

== mlnx-ofed33 ==

* Added: mlnx-ofed33 for Bright 7.1

== pgi ==

* Updated: PGI compiler to version 16.5