On Thu, Aug 19, 2010 at 12:58 PM, Christopher Browne
<cbbrowne at ca.afilias.info> wrote:

Just a data point to add here.  When I had 50k+ objects on my source
database, only 1k or so of which were under slony replication and in
the same set, each lock table issued by slony 1.2.x would take about
45 to 60 seconds.  I'd watch what was going on in pg_stat_activity and
see lots of references to all the tables that were NOT in the
replication set going by in whatever function it was slony was using
to lock its tables.  This meant that doing ddl on the source database
took about 60*1000 seconds or around 16 hours.

Moving those extra 50k+ objects to a destination slony member removed
the long wait and now ddl changes now happen in seconds, or at most
minutes.