i think i got it now. this was a missing piece of the puzzle for me. if the db remains in the availability group then it continues to grow (no truncation) waiting for the AG to come back online. so the proper protocol in a true DR is to remove the db from the AG. then after recovery, backup /restore back to the primary site.
but for a DR test, which will only last a few hours at most, and not a huge amount of transaction during the test, i could set the DR replica to synchronous, do the test, then failback and reset it back to asynchronous. we won't be performance testing the DR site, just validating that failover works and functional tests.
do i have that correct?
Sorta yes, sorta no.
The AG itself will not go offline. It will move the primary to the DR site. The former primary (now secondary) will be offline for whatever reason (crash / DR exercise / maintenance).
If you're only doing a DR test, then nothing goes offline. All you need to do is manually shift the primary to the DR site (making the primary the secondary) then switch it back when you're done. I wouldn't worry about changing the synchronous mode unless you have a real need for your tests to be available ASAP on the regular primary. And if this is production, make sure your tests are using real data that you want left in the system or test accounts that are ignored in all your financial records.
EDIT: If this DR test is mainly about connectivity and the tests are few and small (data wise) then don't bother changing the synch mode at all. There won't be enough data to commit once you change over and most likely it will already be committed since the primary will still be up, albeit as a secondary. Unless... are you saying you plan on taking down the primary server completely as part of this test?? :END EDIT
If this was a real disaster, then the former primary would probably be down. I'm assuming the server or data center would be completely unavailable. But that doesn't mean the AG itself will be offline. It will still be functioning on the available servers. Or should be if it was set up in such a way that it could survive a data center crash (i.e., crossing multiple data centers).
Does that make sense?
Brandie Tarvin, MCITP Database AdministratorLiveJournal Blog: http://brandietarvin.livejournal.com/[/url]On LinkedIn!, Google+, and Twitter.Freelance Writer: ShadowrunLatchkeys: Nevermore, Latchkeys: The Bootleg War, and Latchkeys: Roscoes in the Night are now available on Nook and Kindle.