Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 6.0-16 (released 2012-09-14)
buildmaster: Renamed default burn config to default-destructive. (Only affects new installs.)
buildmaster: allow multiple drives to be selected for non-raid disklayouts
cluster-tools: Updated dependencies-rev.log, revision of package changed, on which cluster-tools depends.
cmburn: The HPL test would not work when there was no cmsupport user. It now temporarily adds a local user to run the test.
cmburn: When a node did not PXE boot after completing post-install burn tests, the burn would never complete. It will now retry the reboot to PXE.
cm-config-cm: Update cm-config-cm dependency check, to exclude bright-installer, buildmatic-${cmversion}-stable, buildmatic-common
cm-config-syslog: Changes in the default rsyslog.conf config file -- removed the log message rate limitting for rsyslog daemons which have it enabled by default (which is set by default to 200 messages per 5 seconds per process -- everything above this limit was being dropped by rsyslog). The rsyslog.conf is *not* updated automatically on existing cluster. It is advised to remove the rate limitting manually for existing clusters as CMDaemon running in debug mode tends to log at peak moments far more than 200 message per second.
cmdaemon: Running the cmd binary with the '-t' option no longer causes a crash on some occasions. Also, the result of using the -t option will now be output to the console without the user having to explicitly use the additional -n flag.
cmdaemon: Fix spurious wake ups in monitoring code
cmdaemon: reduce long startup delay when a lot of nodes are down
cmdaemon: increase master IPs soap call timeout, to prevent error
cmdaemon: keep mysql dump before a cmha clone
cmdaemon: fix drain for torque
cmdaemon: better handling of errors in xml
cmdaemon: fix crash programrunner if free socket limit is reached
cmdaemon: more informative restart requird messages
cmdaemon: Adding functionality to force reinitialization of metriccollections on compute nodes
cmdaemon: Added burn log and burn status tokens to read-only profile. Existing installs will also be modified.
cmdaemon: Adding functionality that allows one to write to the eventbucket (head node or compute node) to force reinitialization of metriccollections:
echo "REINITIALIZE-METRICCOLLECTIONS" > /var/spool/cmd/eventbucket
or if you want to reinitialize only a particular device of which the metriccollection to be reinitialized is sampled on head node:
echo "REINITIALIZE-METRICCOLLECTIONS:genericdevice01" > /var/spool/cmd/eventbucket
cmdaemon: append search domains when using dhcp, before manual settings were overwritten
cmdaemon: Added advanced config option to add external submit hosts to workload managers
cmdaemon: fix occasional crash in cmsh burn start command
cmdaemon: Removed round bracket characters around the timestamp when logging to console.
cmdaemon: Added sorting of the currently running tests as reported in burn status.
cmdaemon: Added an include directive into the named.conf file into the end of the 'options' section. The new include directive includes a file named.conf.global.options.include. The mentioned included file is empty by default, but can be used by the users to add custom named options to the options section without having to freeze the generation of the whole named.conf config file. The included file is created while generating the content for the named.conf file.
cmgui: add log button to burn overview
cmgui: sort burnlog
manuals: improved: single head to HA conversion notes
manuals: added: double ssh -X session avoidance with cmgui
manuals: added: creation of pem/key certs via openssl
manuals: added: add-on installation notes
manuals: added: miimon note
manuals: added: oomkiller healthcheck
manuals: added: reinstalling after wiping the hard drive
manuals: added boot role notes
manuals: fixes and edits for usage of install-bright command
manuals: added: modifying profile example
manuals: cuda 4.2 included, clarity improvement about booting/compiling (rt2761)
netcdf | Since netCDF 4.2 the C++ libraries and headers are no longer provided by Unidata with the core netCDF package (they are distributed separately by Unidata). This fix changes the Bright test script to reflect this fact by now trying to compile and run a C netCDF program using C netCDF library instead of a C++ netCDF program using C++ libraries (which are now missing in this package).
node-installer | During burn the installmode would temporarily be forced to AUTO, both when switching from pre-install to post-install tests, and after completing the burn. This was changed so that the configured installmode for a node is used. In addition, it is now possible to add a new tag AUTO|FULL|MAIN|NOSYNC to the burn configuration, as a child of the tag. The specified installmode will be used when the burn completes.
node-installer | In some cases the motd file on a node would not get restored when the burn completed. We now backup and restore the original motd.
node-installer | Improved stability when a node being installed is configured in a incorrect way.