Last time we installed a new callback for a queue. The next day this callback failed. The side effect was that the RDBMS did not make automatic calls of callbacks for messages enqueued after the message, for which the callback failed.
We have no entry in the alert log nor traces, so this was quite weird.
The solution is somewhat trivial. One needs to improve failing callback - which is obvious, because if not, we will hit the same situation again. The second thing is to dequeue all the not consumed messages and possibly enqueue them again (or rerun the processes, which register such messages).
More details one may find at MOS 1265138.1
No comments:
Post a Comment