4 Replies Latest reply on Mar 10, 2006 10:46 AM by fallow

    GraphSession.deleteProcessInstance taking over 10 minutes pe

    fallow Newbie

      Is anyone else having performance problems related to the deleteProcessInstance call? We are having extremely poor performance on this call. In the order of 10-15 minutes per delete.

      We are running against an Oracle 9i DB. The DB has 2 CPUs dedicated to our schema and plenty of memory. Everything else is running well enough.

      We have tracked down the call that is taking forever. Its the 'delete from JBPM_LOG where TOKEN_=?' call.

      This table has 2.5 million entries in it and indexes on the ID_ and TOKEN_ columns. We add about 2000 instances per day to the system. At the current delete rate we cannot delete finalized instances nearly as fast as we put them in.

      Has anyone else seen this problem?

      Do we really need to keep these log entries around? Can we bulk delete them based on date for instance? If we do a bulk delete by date will this affect the data integrity of the other tables?

      Is there a way to disable creating these log entries?

      We are currently running JBPM 3.0.1.

      Any help would be appreciated.

      Thanks,

      Fallow