Improve Notification table clean up.

Root Cause:

IBM Cognos BI supports batch processing for a variety of tasks and objects. A common example would be a scheduled job containing report executions. The monitor service stores the job and report execution commands in a Task Queue. Commands on the task queue are run as the system has capacity to do so. There is a balance to be struck between the Cognos installation infrastructure and the rate at which background processing occurs, or put another way your installation hardware will have a maximum capacity for background processing that you cannot exceed.

You will notice that if your scheduled execution rate exceeds your system capacity then the Task Queue will start to fill up. If you open the current activities view in Cognos Administration UI, it will have many Pending report executions. Stopping the scheduled activities using the upcoming activities view is one way to solve this issue and let the backlog of reports be processed however this is quite tedious. Another common way that some customers have chosen to alleviate this issue is the NC_DROP procedure. This procedure involves dropping all of the tables in the notification database that are prefixed NC_, and when the system restarts, the tables are rebuilt and repopulated. However, this procedure has a drawback that all of the schedules are re-synchronized to the NC tables which can take some hours on a large, distributed Cognos BI system.

Clean up Candidates:

Criteria: SELECT TOP "+_maxrecords+" t1.TASK_ID from NC_TSE_STATE_MAP as t1, NC_TASK_QUEUE as t2 where t1.TASK_ID = t2.TASK_ID and t1.STATUS >= 20 and ((t2.DATE_ENTERED + "+olderThanTime+") <= (SELECT MAX(DATE_ENTERED) FROM NC_TASK_QUEUE))

Normally, finished tasks get cleaned up one by one by a background thread. On busy servers that thread does not get a chance to clean finished items. This code is for cleaning such candidates in one go in a single transaction.

On busy servers the entries may look like this:

e.g.

50 candidate for cleanup found 123156423, 1232454132, 123213213, .....

Advanced Properties:

event.finished.check.active = true or false // default is true

The property is a flag to enable or disable the finished task cleanup process it is true by default and can be disabled by setting this value to false.

event.finished.check.threshold = 100 // default is 10 finished tasks in older than 24 hours in nc tables

The property defines the maximum number of finished entries in the NC tables selected for removal.

event.finished.check.interval = 86400 // default is 3600 i.e. 1 hour but ideally it should be 24 hours 24*3600

This property defines the repeat schedule for the processing, the tasks which are finished over 24 hours ago would be the candidates for the bulk cleanup process.

event.finished.check.time.threshold = 3600000 // default 86400000 i.e. 1 day

The clean up candidate criteria for finished tasks older than threshold (default 1 day)

Tables affected / entity map.parent are only:

NC_TASK_ANCESTOR_STOREIDS

NC_TASK_HISTORY_DETAIL

NC_TASK_QUEUE

NC_TSE_STATE_MAP

NC_TASK_PROPERTY