Posted: April 05, 2014 14:01 by Cazav
I'm using Processing Pattern release 12.1.0. I've found the following issue:
1 - Two storage enabled nodes joins the processing cluster.
2 - A client (a non storage enabled node) joins the cluster and sends tasks to the cluster. They are dispathed using the round robin dispatcher.
3 - An storage enabled node leaves the cluster (the JVM is suddenly shutdown).
4 - The client continues sending tasks. These tasks are still being dispatched to all the nodes (the 2 storage enabled ones) despite having only one active.
The problem seems to be related with the taskProcessorMediatorCache not being updated on the leaving task processor so the Map<TaskProcessorMediatorKey, TaskProcessorMediator> taskProcessorMediators parameter received has the old list.
Thank you very much.