We have a production database that contains data which is easily
recreated at runtime. I'm considering upgrading this to 9.0 beta1 to
get some experience with the new hot standby system on a server that
is under medium to heavy load.
Obviously, being a production database, it's inconvenient if this
fails (but not catastrophic for us, which is why I want to try it).
If partial updates are enabled and the master fails, I'd still like to
fail over to one of the slaves. If the last log file was not shipped,
but part of the queries for it were replicated, what happens?
What happens if I have two slave servers (A and B) and I want to fail
over to slave A and have it start shipping log files to slave B but B
has more queries applied to it than A? I assume in this case I would
instead want to fail over to B and ship to A. How would I know which
server to fail over to?
Thanks,
Bryan