[UIMA-1665] UIMA AS aggregate not handling timeouts correctly for parallel delegates

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Fix Version/s: 2.3AS
• Component/s: Async Scaleout
• Assignee: Unassigned
• Reporter: Jaroslaw Cwiklik
• Created: 13/Nov/09 17:57
• Updated: 16/Nov/09 17:31
• Resolved: 16/Nov/09 17:31

Description

Burn discovered a bug in UIMA AS aggregate while handling timeouts from delegates in a parallel step. A delegate timeout that is subsequently disabled (due to error handling setting action=disable), causes error handling code to (incorrectly) double increment the count of delegates that responded. The count is maintained to determine how many delegates in a parallel step responded and if it is ok to allow the CAS to continue with the flow. This counter is incremented when a reply comes or there is a timeout. When the counter reaches the expected number, the CAS is allowed to continue on with flow. The double increment causes a CAS to continue in the flow prematurely.