On 11/2/2005 6:41 AM, Andrew Sullivan wrote:

> On Tue, Nov 01, 2005 at 06:50:05PM -0500, Rod Taylor wrote:
>> You can move around the master for a specific set but as Chris reminds
>> me failover itself is a global operation (performed only when the
>> original DB is no longer available).
> 
> Yes, but as Chris also noted, if one database has failed in some way
> on a cluster such that you want to failover, that would seem to
> entail that the rest of them have too.  For any case where that isn't
> true, you can still do MOVE SET.   (Actually, I can think of a case
> where it might not be.  If you're using tablespaces to put each of
> these on a disk, and you have a disk-controller failure for just one
> disk pair, say, then you might have a case.  But that's already way
> more administrative burden than I assume the OP was going for.)

I just wonder what the value in that MOVE SET "on corruption" is. We can 
only talk about logical corruption here, the case where the application 
behaves like the elephant in the china shop. MOVE SET ensures that all 
those "changes" are replicated before the subscriber takes over. Doesn't 
that mean, replicate the corruption before switching to the subscriber?


Jan

-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #