OLAT-5799: [merged from 630public branch] countConnectedUsers made more stable: catch RejectedExecutionException and issue a log.error, leaving the user count at the wrong figure for another 30 seconds. Avoids a RedScreen at DMZ/Login in case the Threadpool is full.
OLAT-5799: countConnectedUsers made more stable: catch RejectedExecutionException and issue a log.error, leaving the user count at the wrong figure for another 30 seconds. Avoids a RedScreen at DMZ/Login in case the Threadpool is full.
OLAT-4930: [merged from 6_3_public to HEAD]: avoid garbage in the filesystem due to loading a course which was deleted while we were indexing (between getting the list of repositoryentries and processing the deleted course) - plus avoid a nested doinsync which also happens because of this
PLUS:
commit the session before looping through the individual entries - assures that even before the first intermediateCommit() no entity is taken from hibernate session cache (as that cache still holds all entities at loop start time). The session cache is released after commit - which happens every time a document is added to the indexer - hence rather quickly. But if an entity would be deleted before the first indexed one, it would not have been detected as deleted using session.get()
added to RepositoryIndexer.java and modified 2 java files
OLAT-4930: [merged from 6-3-public branch] safer closing of streams - small chance that this is related to OLAT-4930 where the CourseConfig.xml might prevent the deletion of a course dir
OLAT-5784: previous checkins re duration added to log.info in deleteCourse and //o_clusterOK added to the newly added synchronized and the one which had o_clusterNOK have erroneously been marked with OLAT-4930
OLAT-4930: commit the session before looping through the individual entries - assures that even before the first intermediateCommit() no entity is taken from hibernate session cache (as that cache still holds all entities at loop start time). The session cache is released after commit - which happens every time a document is added to the indexer - hence rather quickly. But if an entity would be deleted before the first indexed one, it would not have been detected as deleted using session.get()
added to RepositoryIndexer.java and modified 2 java files
OLAT-4930: use the same pattern as in RepositoryIndexer which is: reload the group/identity before indexing it - since the while loop could take rather long and the group or identity could have been deleted already
added to GroupIndexer.java and modified IdentityIndexer.java
OLAT-4930: avoid garbage in the filesystem due to loading a course which was deleted while we were indexing (between getting the list of repositoryentries and processing the deleted course) - plus avoid a nested doinsync which also happens because of this
Bugfix for OLAT-4651 improvement: throughout 6.3.0-6.3.2 the calendar links were not rendered after a calendar had to be reloaded (after either a server restart or logging into a different node)
RSS