Background:
We use many aggregation, singleton and multi-ton orchestrations similar to the Seroter Round Robin method described here (BizTalk 2009).
All these types of orchestration have rather arbitrary exit or continuation points (for aggregations), usually determined by a timer - that is, if Orch no longer received messages within X minutes, then proceed with batch processing, and if more minutes have elapsed, and more messages no. (We also exit Single / N-Tons due to concerns about subscribing to a singleton period for a specific period).
As far as we tried to soften Zombies, for example, by starting any continuation processing in an asynchronous reorganized orchestration, there is always a point of weakness when a “well” synchronized message can cause zombies. (i.e. receiving more incoming messages correlated with "already completed" forms of orchestration),
If a message causes zombies in one of the signatures, the message does not seem to apply to other subscribers (i.e., orchs completely separated from the zombie orchestration), that is, the message that causes the zombies is not processed.
Question
, - , , , "" , . ( ..)
, API BizTalk SQL- MsgBoxDB.