Hi all -

Have seen tons of great stuff discussing synchronous (SAFETY ON) and asynchronous mirroring, including an excellent whitepaper on MS' site. Still, I have not seen the following two questions discussed anywhere. Hoping some of you have some insight to share.

1. With async mirroring the principal database does NOT wait for the mirror to commit the transaction before it (the principal) commits. So, what happens if for some reason the mirror ends up rejecting that transaction???? Wouldn't that mean the mirror no longer reflects the principal???? If that's the case, how are enterprises using asych controlling this exposure?

2. Again w/ asynch mirroring: The SEND_QUEUE (on the principal) holds any backlog of transactions that cannot be sent to the mirror as quickly as they are coming into the principal. This, from my investigations, represents the "possible data loss" that could occur if the principal goes down and you need to failover to the mirror. Easy enough, and there are system queries that one can use to obtain metrics that discuss how much work this might be (i.e., are you losing the last .5 seconds or work? Last 5 seconds? Last 5 minutes? hours?). Of course, this will change from site-to-site and workload-to-workload. Still, I'd love to hear folks talking about what--in their environment--the "possible lost data" amounts to, as clearly the decision to use/not use asynch (aka HIGH-PERFORMANCE) mirroring would be impacted by how long before failover you might have to go back to redo work. Can anyone provide any insight into this?

TIA