Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 7.2-14
== cmdaemon ==
* Various improvements to facilitate upgrades to Bright 7.2
* Changed: use XFS by default (instead of ext4) for devicemapper storage backend when Docker role is assigned to a node
* Changed: ceph osd default pool size is now set to 3 by default
* Changed: job metrics: CPU related metrics are now cumulative
* Changed: partition timeservers are no longer used for normal nodes
* Changed: it is now possible to set ipmi power control for switches
* Changed: disallow setting hostname 'master' for compute nodes
* Added: adv. config. option ExtraRsyncOptions to specify rsync options per category or node
* Added: adv. config. option SSLServerMethod to control the SSL server method
* Added: for a provisioning role, ability to specify all images are on a local disk except those listed as on a shared storage
* Added: JSON call to get user information based on cookie
* Added: drac0 to the list of valid options for device powercontrol
* Added: script to get raw monitoring data directly from the database
* Added: standalone service cm-nfs-checker to cache and query good/stale/broken NFS
* Added: in pythoncm, batch functions for devices
* Added: support for ELK integration in cmdaemon
* Improved: pythoncm documentation
* Improved: check if /cm/shared is mounted before (re)starting workload manager services
* Improved: handling incomplete HTTP responses in Curl
* Improved: Curl client memory allocation
* Fixed: rare crash on HA setup
* Fixed: cumulative metric data sometimes missing NULL markers
* Fixed: writing CONNECTED_MODE=no if specified for the IB interface
* Fixed: rare crash in getMasterIPs
* Fixed: renaming an image, doesn't rename it's revisions
* Fixed: creating kubernetes certificates for users without policies
* Fixed: an issue with applying kubernetes readonly user policy
* Fixed: in some cases, docker started before its configuration has been updated
* Fixed: dell settings refresh
* Fixed: cmd crash when htdocs contains a directory symlink
* Fixed: when Kubernetes::Node role is removed from a device or a category, the impacted devices are correctly unregistered from Kubernetes cluster
* Fixed: in rare cases, an issue with initialization of services list based on the assigned roles
* Fixed: an issue with cmsh rshell
== node-installer ==
* Added: /scripts/ipoib_mode file to select which IP over IB mode is used for all IB interfaces
* Fixed: node-installer failing to update the PXE configuration file for vnodes
* Fixed: writing IB connected/datagram mode based on the configuration of the network interface
* Fixed: mkinit_rd not keeping initrd.orig for standalone nodes
* Fixed: node-installer/scripts/power did not parse Infiniband BOOTIF defined on cmdline
* Fixed: an issue with restarting burn when using boot-over-IB
== manuals ==
* Added: job metrics
* Added: AllowImageUpdateWithAutoMount, MembershipQueryInterval, JobsSamplingMetricsInterval
* Added: cloning monitoring and extra databases
* Updated: cm-hadoop-maint
* Updated: additional projects
* Updated: hadoop config groups -- customizations
== buildmaster ==
* Changed: new head node installations will come with official puppet rpms
== cluster-tools ==
* Various improvements and bugfixes in cm-docker-setup
* Various improvements and bugfixes in cm-ceph-setup
* cm-kubernetes-setup: set user ownership of .kubeconfig
* Improved: OpenStack setup prechecks and diagnostics for MariaDB server rpms
* Added: support for Cloudera 5.5.2
* Added: support for Kafka 0.9.0.1
* Added: support for Hbase 1.2.0
* Added: support for Zookeeper 3.4.8
* Added: support for Accumulo 1.7.1
* Added: support for Hortonworks 2.4.0
* Added: support for Cloudera 5.6.0
* Added: support for Apache Ignite 1.5.0
* Fixed: MIC nodes have the same SwitchPort pointer as host
== cm-apache-hadoop ==
* Added: support for Hbase 1.2.0
* Added: support for Zookeeper 3.4.8
== cm-apache-hadoop-extras ==
* Added: support for Kafka 0.9.0.1
* Added: support for Accumulo 1.7.1
== cm-cloudera-hadoop ==
* Added: support for Cloudera CDH 5.5.2
* Added: support for Cloudera 5.6.0
== cm-config-http ==
* Added: ELK link in the bright landing page
== cm-curl ==
* Updated: to version 7.47.1
== cm-hortonworks-hadoop ==
* Added: support for Hortonworks 2.4.0
== cm-scale-cluster ==
* Fixed: node cloning
== cmgui ==
* Added: grid plot for job based metrics
* Added: support for Cloudera 5.5.2
* Added: support for Hbase 1.2.0
* Added: support for Zookeeper 3.4.8
* Added: support for Hortonworks 2.4.0
* Added: support for Cloudera 5.6.0
* Added: Yarn HA wizard option in CMGui.
* Improved: etcd, docker, and kubernetes role settings management with cmgui
* Improved: sum job metrics per user
* Fixed: in some cases, changing a parameter, such as IP, can cause cmgui to change it for all visited object of the same kind
* Fixed: an issue with API hash checks
* Fixed: cmgui disconnect due to timeout when handling large job counts
* Fixed: an issue with attaching PDU
* Fixed: an issue with eventviewer window update
* Fixed: an issue with update of new metrics or parameters of a metric not being saved
== cmsub ==
* Changed: remove /cm/local/modulefiles/cmsub/ directory when uninstalling cmsub
== cuda70 ==
* Fixed: verify scripts. Previously when alternative build directory was selected, the scripts tried to run the example GPU programs from the default path. Now the scripts will run the programs from the alternative directory.
== cuda75 ==
* Fixed: verify scripts. Previously when alternative build directory was selected, the scripts tried to run the example GPU programs from the default path. Now the scripts will run the programs from the alternative directory.
== mlnx-ofed32 ==
* Updated: to version 3.2-2.0.0.0.
== node-installer-nfsroot ==
* Updated: glibc package, to resolve getaddrinfo stack-based buffer overflow.
== slurm ==
* Fixed: hwloc integration
== torque-slave ==
* Added: libcgroup-tools as a dependency for torque-client