Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 7.1-19
== cmdaemon ==
* Changed: Use SSL_OP_SINGLE_DH_USE openssl option (CVE-2016-0701)
* Added: support for the striping and mirror for the disksetup when using LVM
* Added: --force option to make it possible to switch compute node HA when power control is not working
* Added: adv. config. option WMStatMaxJobs=4192 to limit number of WLM jobs that are kept in memory when calculating metric data
* Added: cm-clone-monitoring-db.sh helper script for cloning monitoring database
* Added: support for $hostname variable substitution for FSMount
* Added: adv. config. option SendDeviceStatusEventToNode defining if nodes should receive device status events
* Added: adv. config. option GridEngineUnregisterNodeWhenDown to shutdown sgeexecd when CMDaemon powers off or terminates the exechost
* Added: adv. config. option DNS::options_* to configure DNS options and be able to limit recursive DNS queries
* Added: adv. config. option NullFSID to define which fsid will not be added to the exports file and allow setting fsid=0 for an export
* Added: interfaces MACs to the environment created for the scripts run by cmd
* Improved: Send an email to the administrator if compute node HA script has failed
* Improved: load and save OSService configuration to xml file
* Improved: display cmsh sysinfo command total disk space in GB / TB
* Improved: generation of /etc/hosts for nodes which do not share the head node management network
* Improved: prevent imageupdate on nodes that have configured automounter
* Improved: support for cmsh readonly fields
* Improved: memory management of monitoring subsystem
* Improved: configuration overlay role retrieval
* Improved: threading model for cmsh: device pexec
* Fixed: dhcpd config for ScaleMP nodes based on softwareimage revisions using the wrong path
* Fixed: removal of old files in /tftpboot/efi64/pxelinux.cfg/
* Fixed: in some cases, usage of incorrect path for shorewall command for sles11/sles12
* Fixed: an issue with /tftpboot sync to cloud-director
* Fixed: cmsh printing of tribool values
* Fixed: cmsh format --reset command
* Fixed: in some cases, mysql healthcheck failure
* Fixed: automatic detection of default scheduler in cmsh jobs / jobqueue modes
* Fixed: in some cases, head node alias interface added to DHCPDARGS in "/etc/sysconfig/dhcpd"
* Fixed: serial over lan getty enable for systemd inside images
* Fixed: an issue with setting burnconfig xml with cmsh
* Fixed: bringing up shared network interface BOOTIF:alias on compute node HA
* Fixed: starting passsive-services on passive compute node HA
* Fixed: nodes' status not changing to DOWN when rebooted using CMSH
* Fixed: an issue with compute node HA nodes staying passive
* Fixed: detection of number of LSF job processes
* Fixed: memory leak in monitoring info messages when switching active head nodes
* Fixed: memory leak cleaning up remote sessions
* Fixed: LSF accounting file parser
* Fixed: displaying openlava jobs in cmsh and cmgui
* Fixed: in some cases, mysql mirror query failing on incorrect column count
* Fixed: an issue with setting switch ports with cmsh
* Fixed: in some cases, crash of head node cmd if a client connects before the headnode is fully up
* Fixed: cmdaemon checking all services when the status of a single service is requested
* Fixed: an issue with too many event delivery threads if one of the compute nodes is hanging
* Fixed: incorrect visual representation in cmsh when the commit of ceph object has failed
* Fixed: LSF/openlava drain/undrain operations
* Fixed: in some cases, an incorrect entry in the reverse DNS files for a compute node alias interface
* Fixed: in some cases, crash of cm-ceph-setup when rebooting nodes
== node-installer ==
* Added: support for nvme devices in disks script
* Fixed: MTU problems when using and internal network with jumbo frames as flat network in OpenStack. (Requires update of node-installer-slave package in the software image too.)
* Fixed: in some cases, node-installer selecting wrong ip for the head node when multiple networks and bridges are defined
== node-installer-slave ==
* Fixed: MTU problems when using and internal network with jumbo frames as flat network in OpenStack. (Requires update of node-installer package on the head nodes too.)
== manuals ==
* Added: cm-hadoop-maint
== Lmod ==
* Updated: to version 6.0.1.
== buildmaster ==
* support for NVMe devices in head node installer
== cluster-tools ==
* Added: support for Spark 1.5.1
* Added: support for Cloudera 5.3.8
* Added: support for Hortonworks 2.3.2
* Added: support for Storm 0.10.0
* Added: support for Hbase 0.98.16.1
* Added: support for Kafka 0.9.0.0
* Added: support for Cloudera 5.4.9
* Added: support for Spark 1.5.2
* Fixed: adding mic overlays when cm-mic-setup runs in interactive mode
* Fixed: In some cases, cm-ceph-setup waiting indefinitely on 30% progress step
== cm-apache-hadoop ==
* Added: support for Hbase 0.98.16.1
== cm-apache-hadoop-extras ==
* Added: support for Spark 1.5.1
* Added: support for Storm 0.10.0
* Added: support for Kafka 0.9.0.0
* Added: support for Spark 1.5.2
== cm-cloudera-hadoop ==
* Added: support for Cloudera 5.3.8
* Added: support for Cloudera 5.4.9
* Added: support for Cloudera 5.5.1
== cm-hortonworks-hadoop ==
* Added: support for Hortonworks 2.3.2
== cm-openstack-caas ==
* Removed: swap partition from default disk layouts.
* Added: sub command start, shutdown and shelve to os_cluster.
* Added: external network port to all compute nodes by default. Makes deploying OpenStack easier.
* Added: remove as an alternative to delete subcommands.
* Added: help command. (As alternative for -h flags.)
* Fixed: List clusters would show any instance with a floating IP address.
== cm-scale-cluster ==
* Fixed: an issue with starting physical nodes
* Fixed: an issue with starting explicitly requested by a job compute nodes
* Fixed: crash after python-libs update
== cmgui ==
* Increase default timeout for getMonConfs call to 180s (from 60s)
* Changed: default cloud director EBS from 40G to 42G
* Added: support for Hortonworks 2.2.8
* Added: support for Cloudera 5.4.9
* Added: support for Spark 1.5.2
* Added: support for Cloudera 5.5.1
* Added: support for Hortonworks 2.2.9
* Fixed: an issue with using cmgui with Firefox 44.0
== cuda-driver ==
* Updated: nvidia driver to version 352.63
* Updated: to version 352.68
== intel-compilers-2016 ==
* Added: intel-compilers-2016 for Bright 7.1
== intel-mic ==
* Upgrade from 3.5.2 to 3.6.0
* Upgrade from 3.6.0 to 3.6.1
== intel-wired-ethernet-drivers ==
* Updated: e1000e driver to version 3.3.1, ixgbe driver to version 4.2.1, ixgbevf driver to version 3.0.2, i40e driver to version 1.3.47
* Updated: e1000e driver to 3.3.3, igb driver to 5.3.3.5, ixgbe driver to 4.3.13, ixgbevf driver to 3.1.2, i40e driver to 1.3.49, i40evf driver to 1.3.34. Added fm10k driver, version 0.19.3.
== lua-posix ==
* Fixed: removed tests for value of _DEBUG variable. The tests required _DEBUG to be declared to pass the strict check.
== modules ==
* Fixed: previously when starting a new (bash) shell, the MODULEPATH enviroment variable was reset to the default. With this fix the MODULEPATH enviroment variable is not reset to default anymore.
== pgi ==
* Updated: to version 15.10
== slurm ==
* Upgrade to 14.11.11. This version includes a fix a buffer overflow bug that has been discovered in Slurm version 14.11.