Recover and Replication

Just a quickie

If I have a 2 master (asynchronous, symmetric) setup: A and B. The following happens:

1. I do a cold backup of "A", and start it up again
2. A user creates a transaction upon B
3. B's transaction is replicated to A
4. I have corrupt A (severely)
5. I restore from cold and recover A

question: What happens to that transaction from a replication standpoint? That is, the transaction is in my Redo on A, and as I roll forward that transaction is applied on A. But, since I'm rolling forward, and applying redo on A, aren't those transactions now being queued to be applied to B? My guess is they are not, but how does Oracle tell the diff between transactions that are being applied due to replication and those due to recovery?

I guess that leads to another question. How does Oracle avoid the whole circular transactions in general? That is, I make a transaction on A, that transaction is also added to the queue on A, Then the queue is processed, and the transaction is applied to B. Why doesn't (how doesn't) B, now put that same transaction in its queue and try to apply it to A?

Thanks
Jake Received on Tue Jul 16 2002 - 22:46:21 CDT