Re: [rules-users] Weird Problem: WorkItemManager seems to loose WorkItems

Thanks for the suggestion, but since I work in two "worlds" (the process definition and the process instance), I couldn't find one object I could reliably synchronize on. As far as I know, Drools internally syncs on the process object in the "definition world" and on the process instance object in the "instance world" (at least in the parts of Drools I had a look at, i.e. the parts that used to cause concurrency issues for me). And just to be clear, when I say I synchronize on those two objects, I mean I acquire both locks for my critical sections (-> nested synchronized-blocks). My main goal is not to change any of the Drools internals so that I can deliver my code as an add-on to a standard Drools install. Also, I guess it's a lot easier to upgrade to newer versions of Drools if I don't change any internals. After reading a recent thread on concurrency issues of the StatefulKnowledgeSession (http://drools-java-rules-engine.46999.n3.nabble.com/StatefulKnowledgeSess...) I have a theory of what might be going on in my case: Due to the number of process instances I start in parallel it is likely that many of them reach a WorkItemNode at the same time (right now I start the same process multiple times, so the execution times for all instances should be more or less the same). When that happens, I guess it is likely that two WorkItems are created at the same time with the same WorkItemID (since DefaultWorkItemManager doesn't seem to be thread-safe with regards to the ID generation). All WorkItems are stored in a Map with the WorkItemID used as the key, so only one of these WorkItems with the same ID is stored in the Map. And there we are: one WorkItem is "lost" and the corresponding ProcessInstance will never finish. I really hadn't thought about this before because I assumed the usage of ConcurrentHashMap for WorkItem storage in the DefaultWorkItemManager would ensure thread-safety. I'm still testing the solution provided in that thread but so far it looks very nice. Regards Henning Losert -- View this message in context: http://drools-java-rules-engine.46999.n3.nabble.com/Weird-Problem-WorkIte... Sent from the Drools - User mailing list archive at Nabble.com.