Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 6.0-15 (released 2012-10-02)
cluster-tools: prevent sorting of kernel module list
cluster-tools: fixed multiple --arguments specification
cluster-tools: Fixed PBS Pro setup: if master is not selected as a compute node and afterward pbsproclient role is assigned to the master, then pbs_mom would not be started
cluster-tools: Fix PBS Pro: if 'master as compute' option is selected, then 2 virtual nodes for head node are created and jobs are never been scheduled on the head node
cluster-tools: Added macros: CM_SCHEDULE_TRANSFER
cmburn: Changed date format that gets stored when burn completes, to allow easier sorting.
cmdaemon: add fullPowerStatus function to python
cmdaemon: add timeout option to python power calls
cmdaemon: Fixing possible null pointer in snmp queries
cmdaemon: Further improved burn status when transitioning to a next phase. Also indicate when a burn phase will eventually fail because one of the tests already failed but others are still running.
cmdaemon: Fixed PBS Pro cloud re-configuration
cmdaemon: Changed ntp healthcheck. Decreased timeouts so it will not interfere with cmd timeout. Added a extra check if script is unable to determine time difference, if active time server is not the active master
cmdaemon: Changed ntp healthcheck. changed ntp query timeout to 5 seconds
cmdaemon: fixed custom torquemom service name setup
cmdaemon: Fixing possible overflow in select() call
cmdaemon: Modifying smart.conf, so that 'device log contains errors' is not regarded as fatal anymore. The reason for this is that it turned out to generate a lot of false positives
cmdaemon: add drain status to python
cmdaemon: add terminate to python
cmdaemon: Do not put job environment variables in std output file if cmsub+pbspro is used
cmdaemon: possible fix for the CMD_* not set in metric scripts
cmdaemon: add NodesInQueue metric: see ##2860
cmdaemon: Small fix for burn status to correctly report nodes that are not burning.
cmdaemon: Improved the error message which is relatively frequent for newly setup burst-enabled cluster. The error message was extended with a description of a frequently working solution to the issue.
cmdaemon: gzip support for json
cmgui: update burn
cmgui: health view based on jobs (1)
cmgui: job selection to grid wizard
cmgui: add job group to monitoring
cm-openssl: Fixing dead-lock in openssl (http:
intel-cluster-runtime: set RPM macro __os_install_post to %{nil}, to prevent "zipimport.ZipImportError: not a Zip file" bug, for mpitune
intel-mpi: set RPM macro __os_install_post to %{nil}, to prevent "zipimport.ZipImportError: not a Zip file" bug, for mpitune
intel-mpi: Added I_MPI_ROOT enviroment variable to module files.
manuals: improved: provisioning subsystem rewrite and update
manuals: improved: occupationrate explanation
mvapich: Changes in the self-test script -- added a test to check for InfiniBand devices being present before the self test is perform.
mvapich2: Changes in the self-test script -- added a test to check for InfiniBand devices being present before the self test is perform.
netcdf: append LD_LIBRARY_PATH in netcdf module files
node-installer: Added a non-interactive plain text install mode to the node installer. When running in this install mode, the node installer does not make use of the ncurses library . It attempts to perform installation on a best-effort basis.
Best effort means that because this is a noninteractive mode, whenever a regular node installation would REQUIRE a user interaction in order to proceed, when running in non-interactive plain text mode the installer will simply abort installation. When running in the non-interactive mode the node installer logs all output to the screen. The non-interactive plain text mode is now also the default install mode for all cloud nodes.
It's possible to force it on a non-cloud node by adding the following kernel parameter to the software image 'bcmplaintext=1'.
node-installer: Added a script named 'debug' to the '/cm/node-installer/scripts' directory. The script can be used to easily toggle the debug mode of the Node Installer. The script toggles the debug mode by editting the 'debug' directive inside of the '/cm/node-installer/scripts/node-installer.conf' file.
node-installer: Increased the verbosity of the BMC configuration error message.
node-installer-slave: Updated dependencies-rev.log, revision of package changed, on which node-installer-slave depends.
node-installer-slave: Updated dependencies-rev.log, revision of package changed, on which node-installer-slave depends.
node-installer-slave: Updated kernel-update.log, kernel dist package got updated
pbspro: fixed pbspro mom start on head node