> But since all tables got locked, even for readers, while the subscription
> took place, a reinitialization of the slony cluster was pretty disruptive"

Still though, Slony at some point must have a lock on all structures and
it cannot give up locks as it goes. As far as I know the order that
Slony runs through the structures is not predictable -- escalating locks
is always a bad idea. 

Anybody relying on access to these tables while Slony is building them
will have pretty big problems at some point. Is there a better way to
allow Slony to do its thing without blocking?


Perhaps by creating new structures, loading up the dataset, dropping the
original and renaming the new one?

--