Base Command Manager / Bright Cluster Manager Release Notes

Release notes for Bright 6.0-30 (released 2013-03-27)

== cmdaemon ==

* Added: turn LSF/openlava host into a submit host if slots=0 for the client role
* Added: cmsh device list now accepts -n / -r / -g / options
* Added: advanced configuration directive for snmp large fsset: SnmpLargeFdSet=0
* Added: allow for node[001-010] style formatting in cmsh
* Added: cmsh events follow command
* Improved: device status logging in events.log
* Improved: cache device status in cmsh, such that they can be viewed with device list
* Improved: reload named service when no zone files were deleted, otherwise restart
* Improved: set state to PENDING for all cloud nodes after power on
* Improved: information about all nodes being unnecessary sent to all nodes when torque/pbsproclient is assigned
* Improved: keep alive events for node installer
* Improved: audit log for provisioning calls
* Fixed: using slots defined for sgeclient role
* Fixed: ldap config being written while still negotiating failover
* Fixed: commit with --wait option for softwareimages not always working properly
* Fixed: clean up powersave text when performing power operations manually
* Fixed: disabled prejob checks still being run
* Fixed: LSF queues configuration in some cases being saved incorrectly
* Fixed: parsing LSF 8.x job memory size
* Fixed: setting additional LSF queue parameters
* Fixed: cleaning autogenerated section in /etc/pam.d/sshd
* Fixed: occasional cmdaemon crash when latest monitoring data is being picked up from passive master
* Fixed: slurm drain parsing
* Fixed: software image not marked as modified after using up/down commands for a kernel module
* Fixed: openvpn not being restarted if it was automatically stopped earlier because it was no longer needed
* Fixed: possible gzip memory leak
* Fixed: cmsh pbs wall clock check
* Fixed: cmsh appearing to hang when long power commands are given
* Fixed: torque cloud prolog not being added when a cloud-provider is added
* Fixed: dhcpd and slurmd being restarted on configuration change when they are stopped from cmsh/cmgui
* Fixed: jobs parsing when alternative torque service name is configured
* Fixed: incomplete information message for failed torque health checks
* Fixed: python doc: pythoncm.Network.NetworkType.INTERNAL --> pythoncm.NetworkType.INTERNAL

== node-installer ==

* Added: support for root on Lustre.
* Added: feature to, optionally, sync kernel modules from an image into the nfs root during initrd creation.
* Added: a work around for a possible IPMI configuration problem on some hardware.

== manuals ==

* fixed: shutdown of cloud nodes stops the cloud instances without terminating them

== buildmaster ==

* Improved progress indication on screens during installation
* Check for head node interface on all bootable networks
* Check for head node interface on the internalnet if it is not the management network
* Show drives that are less than 1G in install drive selection
* Update of rhel6 X packages selections, for update to rhel6u4
* Nameserver validation when adding additional nameservers
* Enforce 2.2TB limit for boot devices only for RHEL5 based distributions
* Renamed hwloc rpm package to cm-hwloc
* Added gpu sample to head node monitoring configuration
* Removed 3w-xxxx and 3w-9xxx from the list of default kernel modules for the software image

== cluster-tools ==

* cm-create-image: -l option to provide optional resolv.conf to be used inside image for yum/zypper to work
* cm-register-node: update node certificate autosign days to 36500
* Set user message in power save mode
* Add startup section to drbd.conf to prevent drbd from waiting for user input at startup/boot

== cm-busybox ==

* Changed the busybox modprobe setup so that it respects modprobe.conf.

== cmburn ==

* Cope with different behaviour of symlinks in RHEL6

== cmgui ==

* Collapse tree rack/nodegroup/chassis items by default when connecting to a new cluster
* Added extra EC2 types to price overview
* Added paging to burn overview
* Added paging to cluster health
* Improved switch port overview
* Turn off auto monitoring tree fill by default
* Fixed: issue with metric being removed from rackview
* Fixed: icon spinning for monitoring window
* Fixed: extra tunnel interface for the headnode being added even if one already exists

== libpam_wlm ==

* Proper /etc/security/pam_bright.d/pam_commands.conf

== mlnx-ofed ==

* Log more info about packages that are removed.

== qlgc-ofed ==

* Log more info about packages that are removed.
* Changed install config kernel module config, for rhel5 and sles11, for update from 7.1.0.0.58 to 7.1.1.0.25
* Update from 7.1.0.0.58 to 7.1.1.0.25. Changed install config for rhel6.

== slurm ==

* Added /cm/shared/apps/slurm/current/lib64 to LD_LIBRARY_PATH in the module file