Released 14.12.2011.
Highlights
We have improved the generation of our indices, which primarily should speed up the generation of the index generated for the snapshot harvests. Furthermore, we have made some optimizations to the GUI. Finally, a lot of database cleanup that was handled properly, has now been taken care of.
Upgrade-notes
Be sure to run the dk.netarkivet.harvester.tools.HarvestdatabaseUpdateApplication tool after installation of the new release, but before starting the Netarchivesuite system. See the additional tools manual for further details.
Note that the harvesting setting:
<sendStatusDelay>60</sendStatusDelay>
has been replaced by the two settings
<sendReadyInterval>30</sendReadyInterval> <sendReadyDelay>1000</sendReadyDelay>
See HarvesterSettings.java and Harvester settings.xml for details.
Furthermore a number of new archiver settings has been added in relation to the improved indexing-mechanism:
settings.archive.indexserver.requestdir
settings.archive.indexserver.maxclients
settings.archive.indexserver.maxthreads
settings.archive.indexserver.checkinterval
settings.archive.indexserver.indexingtimeout
settings.archive.indexserver.listeningcheckinterval
And added are also new settings related to the database pool mechanism for the archive database (similar to the one for the Harvest Database):
settings.archive.admin.database.pool.minSize
settings.archive.admin.database.pool.maxSize
settings.archive.admin.database.pool.acquireInc
settings.archive.admin.database.pool.maxStm
settings.archive.admin.database.pool.maxStmPerConn
settings.archive.admin.database.pool.idleConnTestPeriod
settings.archive.admin.database.pool.idleConnTestQuery
settings.archive.admin.database.pool.idleConnTestOnCheckin
See ArchiveSettings.java and archive settings.xml for details about these new archive settings.
For those deploying the software to Windows servers, note that we have now disabled Console-logging on this category of servers, as it occasionally results in hanging bitarchive applications necessitationg a restart of the applications. See issue NAS-1993.
Note that we have added notifications in the HarvestJobManager, if it thinks it is skipping a scheduling, because the previous scheduling is still running (NAS-1968). A known problem with this is that the harvestJobManager will send out lots of notifications during the scheduling of snapshot-harvests. (NAS-2005).
Known problems and work-arounds for these.
- NAS-2030): If your NAS installation includes two Derby database servers, they currently write to the same database log ($INSTALLDIR/derby.log) and use the same
properties file ($INSTALLDIR/derby.properties). The fix is to alter the derby.system.home for one of datatases, e.g $INSTALLDIR/adminDB as follows
after having shut down all apps using the database:mkdir $INSTALLDIR/adminserverdir mv $INSTALLDIR/adminDB $INSTALLDIR/adminserverdir/ add -Dderby.system.home=$INSTALLDIR/adminserverdir after -Xmx1936m in the scripts $INSTALLDIR/conf/kill_external_admin_database.sh and $INSTALLDIR/conf/start_external_admin_database.sh
- The adminDB now uses the same pool mechanism as the harvestDB, i.e. c3p0. It seems that the defaults causes access delays in the database access. This can be fixed by overriding two of the default pool settings (idleConnTestOnCheckin, idleConnTestQuery) in the deploy configuration:
<archive> <admin> <database> <dir>adminDB</dir> <pool> <idleConnTestQuery>SELECT COUNT(*) FROM replica</idleConnTestQuery> <idleConnTestOnCheckin>false</idleConnTestOnCheckin> </pool> </database>