Michael Paquier wrote:
> On Fri, Feb 7, 2014 at 3:02 PM, James Sewell <james.sewell@lisasoft.com> wrote:
>
>> Node A could get ahead even if it has been shut down cleanly BEFORE the promotion?
>> I'd always assumed if I shut down the master the slave would be at the same point after shutdown
>> - is this incorrect?
>
> Yes and no. A node will wait at shutdown until all the *connected* slaves have flushed necessary WALs
> to disk. But if the slave is not connected at the moment of shutdown, it might not be the case.
Even if the slave is connected, there was a bug:
http://www.postgresql.org/message-id/E1UrWwA-0004LR-D2@gemulon.postgresql.org
So you need at least 9.3.0, 9.2.5 or 9.1.10 for this to work properly.
Yours,
Laurenz Albe