Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 6.1-6 (released 2013-07-11)
== cmdaemon ==
* Drain overview token added to the readonly profile
* GET_CATEGORY_TOKEN added to the POWER profile
* Changed the default base address for private clouds from 10.42.0.0 to 10.0.0.0
* Added: GPU healthchecks
* Added: OpenSSHportOnCloudDirectorExternalInterface advanced configiuration option managing SSH access to the cloud director
* Added: automatically allocate/release staticIP when starting a VPC cloud director
* Added: dynamically add, remove, or re-assign subnets to/from a private cloud
* Added: job submission CMDaemon API usage examples for each workload manager
* Added: cm-import-image.py python examples which demonstrating how to use Python API to import a software image from other clusters
* Added: static IPs for cloud nodes inside a VPC
* Added: static IPs for cloud directors in EC2-Classic
* Added: allocate and release Static IPs from within CMSH
* Added: support for spot instances inside private clouds
* Added: support for VPC with custom netmask bits and base addresses.
* Added: support for custom SLURM partition parameters
* Improved: creating a cloudnode in cmsh will automatically set the cloud network for eth0
* Improved: JSON parsing for list of objects from cmgui (ProvisioningRole->nodegroups)
* Improved: fill in new properties of workload managers jobs
* Improved: support for third-part VPN connections to AWS
* Fixed: nameserver views for all subnets that belong to a cloud directors VPC
* Fixed: repetitive emails being sent if failover could not be completed
* Fixed: possible crash in JSON calls on bad input
* Fixed: cmdaemon message when removing a cloud provider which is in use
* Fixed: default prolog script for new lsf and openlava queues not being set
* Fixed: in some cases ec2subnetid of a network not being updated when the subnet is added to VPC.
* Fixed: cloud director dependents node group being cloned when not needed
* Fixed: JSON object conversion
* Fixed: SLURM job submission via CMDaemon API if sbatch uses stderr
* Fixed: cmsh: device foreach -s in some cases not properly working
* Fixed: conman being started if SOL is enabled in the images but no nodes have ipmi/ilo
* Fixed: cmdaemon accepting empty cloud provider names
* Fixed: rare crash during device commit
* Fixed: catch unknown exception which in some cases can crash cmd
* Fixed: cmdaemon allowing multiple identical subnets to be assigned to a private cloud
* Fixed: In some cases cmdaemon not being able to use EC2 accountIDs with dashes
* Fixed: SGE job submission API
* Fixed: possible crash in eventbucket
* Fixed: in some cases incorrect netmask openvpn ifconfig-push for master ccd file
* Fixed: in some cases pbsmom config being updated before nameserver update
* Fixed: config sum cache, which in some cases can lead to false 'The file has been overwritten and a backup has been stored in' events
* Fixed: in some cases node state shortly going down after callinginit
* Fixed: in some cases pbsmon config not being updated on the compute nodes
* Fixed: job submission via CMDaemon API not handling special cases of PBS Pro job submission
* Fixed: job submission via CMDaemon API not handling special cases of LSF and openlava job submission
* Fixed: dhcpd conf file generation according to recommended settings from ScaleMP, using vSMP node's IP and setting PXE path based on VCI
* Fixed: cloning of objects between two clusters form pythoncm
* Fixed: inconsistency in the names of some of the private cloud related fields of network objects
* Fixed: cmsh name resolve issue when the name has been changed by another cmsh/cmgui session
* Fixed: cloud node termination via pythoncm
* Fixed: cmsh tab completion in the private cloud object mode.
* Fixed: duplicate lines in shorewall masq file
* Fixed: pbspro service from being registered as stopped when start has timed out after a config update
* Fixed: PBS Pro in cloud: use IPs in pbs.conf instead of hostname on cloud nodes
* Fixed: openvpn being run on cloud that have no tun interface
* Fixed: cloud-vpc zone not being added to the head node nameserver configuration
* Fixed: tab completion for availability zones in cloud settings
== node-installer ==
* Fixed: disks script quitting if the hardware vendor and custom asserts fail on the first 'found' device
== manuals ==
* Added: CPU Scaling Governors
* Improved: openlava documentation
* Updated: cuda changes
* Updated: SLURM power saving allowed
* Updated: LSF root and var directories
* Updated: provisioning node parameters
* Fixed: description of clear command
* Fixed: openssl path
== buildmaster ==
* renamed prolog-clouds-setup call to prolog.clouds.setup to maintain 6.0->6.1 upgradability
* Improved: send a notification to the admin email about firstboot failure
* cmfirstboot: log to /var/log/cmfirstboot.log, important messages also to syslogd
== cluster-tools ==
* cm-mic-setup: use partition management network if host's one is not defined
* set powerSavingAllowed property for category client role when power save is disabled or enabled with wlm-setup
* check for powerSavingAllowed property of nodes and node category when performing power operations
* Added default VPC support to cloud-setup
* cloud-setup: FSexports inside individual cloud directors, instead of the cloud director category. This is needed to support mixed EC2-classic/EC2-VPC accounts.
* Improved cloud-setup-private-cloud script.
* cloud-check script: improved support for clusters with private clouds (VPCs)
== cmgui ==
* Improved display of MIC cards in rackview
* Improved reinstall node: power reset node when cmd is not running on the node
* Cloud Static IPs management
* Fixed: adding/cloning cloud nodes in cloud overview
* Fixed: removing a jobqueue incorrectly reporting an internal error
* Fixed: cmdaemon allowing two cloud nodes with the same hostname
* Fixed: cmgui problem on older firefox versions
* Fixed: issue with keystrokes to hidden window on Mac
* Fixed: nodes not being shown as closed, when they are up and closed
== cmsub ==
* Fixed incorrect module description in module file
== cuda-driver ==
* Update from 319.23 to 319.32.
== intel-cluster-checker ==
* Added updated license.
== intel-compilers-2013 ==
* Fixed intel-compiler-64 module file, removed MIC path from PATH variable.
== intel-mic ==
* Allow mic/ofed-mic drivers to be built inside software image
* Upgrade MPSS from 2.1.6720-13 to 2.1.6720.15
== intel-wired-ethernet-drivers ==
* Updated e1000e driver from 2.3.2 to 2.4.14, igb from 4.2.16 to 4.3.0, ixgbe from 3.15.1 to 3.16.1
== mpich ==
* changed rpm description of MPICH packages to reflect also MPI-3
== openmpi ==
* upgrade to 1.6.5
== pbspro ==
* Upgrade PBS Pro 12.1.0 -> 12.1.1 (new version includes 11.2 -> 12.x upgrade fix)
* Fixed creation of prologue link on head nodes for pbs_mom
== pgi ==
* Run PGI makelocalrc in post section of RPM, gcc version on can differ per distro release update.
* Update from 13.5 to 13.6
== qlgc-ofed ==
* Update from QLogicIB-Basic 7.1.1.0.25 to IntelIB-Basic 7.2.0.0.42 (RC)
== torque ==
* Fixed TORQUE authorization issue on a failover cluster