Ole Wendland <ole.wendland at werum.de> writes:
> Ok thank you guys,
>   after really dropping the hole DB it worked. I really hoped that I 
> could avoid the step, but this seemed to be an illusion.

Sorry, yes, it's an illusion to think you can avoid rebuilding.

You don't necessarily need to drop the whole database - the scenario
where that's truly needful is where there was a corruption (e.g. - bad
disk, bad memory, or such, corrupting the database).  

In contrast, if the reason for the failover was a network outage, then
there's no reason to believe that the database is actually corrupted,
and it should suffice to drop replication (e.g. - DROP NODE, or DROP
SCHEMA "_WhaTeVER" CASCADE), and rebuild.

(If there was a hardware corruption, I'd be inclined to go further than
merely dropping the database - I'd want to scrape things down to the
filesystem level, and start over with "mkfs" as one of the first
steps...)

In any case, there is no way to reintegrate the failed master, as it may
have state committed on it that was never replicated, and there's no way
for Slony-I to account for that.  

Perhaps there's none of that, in a particular situation, perhaps one
might imagine it OK to try, but we didn't put this "perhaps" into the
implementation, as we didn't think that "perhaps you don't care about
your data that much" represented a wonderful feature.
-- 
output = ("cbbrowne" "@" "afilias.info")
Christopher Browne
"Bother,"  said Pooh,  "Eeyore, ready  two photon  torpedoes  and lock
phasers on the Heffalump, Piglet, meet me in transporter room three"