Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 6.0-23 (released 2012-11-30)
== cmdaemon ==
* Added GET_METRICCLASS_TOKEN to readonly
* Added GET_OSSERVICE_TOKEN to readonly profile
* Added GET_CMSERVICES_TOKEN to readonly profile
* Added CMDAEMON_FAILOVER_STATUS_TOKEN to readonly profile
* Added adjust the oom value of mysqld in the /proc//oom_adj or /proc//oom_score_adj. Users can configure the oom value by using advanced config parameters, keyword "oomkill.service-name" and the oom number as value. E.g. AdvancedConfig = { "oomkill.mysqld=0"}.
* Added AJAX console backend
* Added JSON checks for not sending password info for readonly profiles
* Added download kernel and ramdisk from Amazon EC2 clouddirector
* Added limiter on number of concurrent cloud kernel download threads
* Added parameters to PBS Pro mom config file for an Amazon EC2 cloud node
* Added set queue property of PBS Pro node
* Optimized pbs exec_hosts parser
* Added software image paths to the exported cmdaemon (XML) configuation file
* Added on demand update of the sysinfo of a node: echo UPDATE.SYSINFO > /var/spool/cmd/eventbucket
* Increased default number of threads for parallel calls
* Improved handling IP and cmd.url in Amazon EC2 cloud setups
* Improved cloud director external IP handling: clear the IP on power off
* Improved display info event messages from node installer
* Improved JSON parsing
* Config file write (passive / provision / boot) node now triggers updates of provisioning system
* Fixed telnet shell in some cases taking 100% cpu
* Fixed a problem which, when soap calls to a provisioning node would fail, could cause the provisioning slots to remain occupied.
* Fixed if an image is already removed, a provisioning request could lead to cmdaemon crash.
* Fixed in some cases incorrect myhostname postfix/main.cf on passive head node
* Fixed in some cases incorrect mydestination postfix/main.cf
* Fixed eventbucket: don't retry commands forever - could fill memory up
* Fixed cmsh refresh not working inside cloudsettings
* Fixed tftpboot update on symlink change
* Fixed possible cmdaemon deadlock: failover node slow + cmd stop
* Fixed extra columns in dumpmetricdata
* Fixed unnecessary creation of tun0 config file in some Amazon EC2 cloud setups
* Fixed an issue when ephemeral storage was selected for nodes in the Amazon EC2 cloud (when the head node is also in the cloud)
* Fixed an issue which can lead to fsexport options not properly set
* Fixed incorrectly reported phase load
* Fixed schedulers healthcheck, which in some cases reported false results
* Fixed an issue which can lead to the dhcp flag for an interface being cleared when initializing cmdaemon configuration from an XML file
* Fixed total/phase time for burn status. Total time would keep on counting even after burn completed. Also, phase time would keep on counting after burn failed. Now both total and phase time stop counting as soon as burn completes or fails.
* Fixed deleting an active healthcheck in some conditions can lead to cmdaemon crash
* Fixed acl_hosts for PBSPro
* Fixed support for ${} in fsmounts
* Fixed cmsh layout after 'history'
* Fixed use consolidated data also for statistical dump
* Fixed IB network controller check
* Fixed netmask for ccd files on Amazon EC2 cloud director
== node-installer ==
* Improved ProgramRunner
* Fixed support for ${} in fsmount
* RHEL-based distros: The node-installer will now check the original (provisioned) /etc/sysconfig/network whether it already contains a section generated by the node installer. If this is the case, node installer will remove this section, before generating a new one. Such section might already exists in this file if the file was obtained from a provisioned node node using 'grabimage'. Additionally, the generated section will now not contain a 'NETWORKING=yes' directive if this directive is already present in the original file (which is the case by default). THIS FIX CHANGES THE FORMAT OF THE NODE-INSTALLER GENERATED COMMENTS MARKING THE BEGINNING AND THE END OF THE NODE-INSTALLER GENERATED SECTION IN THE /etc/sysconfig/network. EXISTING CLUSTERS WHICH ALREADY HAVE THE PREVIOUS STYLE COMMENT IN THEIR /etc/sysconfig/network FILE [DUE TO OBTAINING THE FILE VIA grabimage] MUST HAVE THIS NODE-INSTALLER GENERATED SECTION REMOVED FROM THAT FILE IN THE SOFTWARE IMAGE MANUALLY.
== node-installer-slave ==
* cm-clone-install: do not use -z option for rsync
== manuals ==
* added: uname under chroot gives running kernel
* added: pexec example with -j
* added: bonding on head nodes
* added: provisioninginterface values and head nodes
* added: updating a kernel when BCM OFED stack is being used
* updated: imageupdate_initialize and imageupdate_finalize are no longer run during imageupdate
* added: workaround for libpoppler pdf viewers so that they can do proper text search
== buildmaster ==
* Behaviour of head node installation/drive selection has been changed. By default no drives and disklayout are selected. Upon selection of drive(s), minimal and recommended disk size requirements are checked and if requirements are not met, a one-big-partition layout is chosen automatically
* Added change network interface order button on network interfaces screen of head node installer
* Added force option to chkconfig command in cmfirstboot for sles11.
* Fixed: The bridge and bond interface's ifcfg files will not be properly ignored when creating an image from a node which have those interface defined.
* Fixed X package selection for CENTOS6 and RHEL6
* Fixed X package selection for SL6, removed rhn-setup-gnome package.
* Added /media to default grab and update exclude lists. (Only fixes new installs.)
== cluster-tools ==
* cm-create-image: fixed an issue with cm nsswitch.conf not getting installed or getting overwritten
* Use cluster.passivemaster when setting up shared storage
* Do not mount shared resources on passive for das/drbd filesystems
* Fixed using '-l --nodes=N' parameter in case of PBS Pro cloud jobs
* cmha-setup: provide option (-p) to specify mysql password as an argument for automated failover setups
* Fixed call to validate DAS devices in command line mode
== cm-config-intelcompliance-master ==
* Exclude Dell i386 libs from lib32_counterpart_lib64 test in Intel cluster check configs
== cm-workload ==
* Fixed mpi test of PBS Pro cloud submission
== cmgui ==
* Allow ' in passwords
* Added hourglass icon to workload job operations
* Added generic device model name to graph legend
* Added job filter to grid plot wizard
* Added drilldown view
* Improved tree sort
* Fixed edit bridge interface
* Fixed: only metrics for box-whisker
* Fixed healthcheck color
* Use correct x,y values for top rackview: don't strip blank rows/columns
* Correct default sort order, all ascending
* Make table sort aware of MB
* Delay rack fill until resource tree has been filled
* Fixed occupation rate
* Save custom metric bounds
* Allow forcing bounds for rack view
* Fixed overview power status/control
* Fixed locate in rack when rack is not set
* Fixed cloud director role
* Fixed Amazon EC2 spot price < 0.01 not being displayed properly
== conman ==
* Fixed expect scripts for iLO, if public key authentication is setup for iLO.
== cuda50 ==
* Added OpenCL examples and verify_opencl.sh. OpenCL examples are removed from CUDA 5.0, OpenCL examples copied from CUDA 4.2.
* Fixed OpenCL library reference
* Fixed profiler ini file to point to JRE in the toolkit package.
== cvos-config-intelcompliance-master ==
* Exclude Dell i386 libs from lib32_counterpart_lib64 test in Intel cluster check configs
== finalise-base ==
* Resolve uid/gid conflicts when creating system users
== imb ==
* Fixed IMB binary name in example script
* Patch for IMB declare warnings
== intel-compilers ==
* Update of Intel composer suite from 2011.11 to 2011.12
== pgi ==
* Update of PGI from 12.3 to 12.9
* Update from 12.9 to 12.10
== slurm ==
* Added logrotate.d config for slurmctld
* Added: On compute nodes create the /etc/slurm/cgroup directory and install the example scripts in it, also install the cgroup.conf.example file.
* Fixed location of slurmctld.pid and slurmdbd.pid (slurm user doesn't have permissions to write in /var/run/)
== torque ==
* Fixed cm-launcher (could not find instance-wrapper)