John Sidney-Woollett wrote:

> I'm no expert but have you tried vacuuming the master and slave 
> databases?

I expect it would also be worth looking at the script (I think just in 
CVS HEAD, alas) test_slony_state.pl, which rummages through the nodes 
looking for some troublesome conditions.

I'm actually revising it right now (it's in an editor buffer; should get 
checked in later today...) to actually evaluate the conditions so that 
it alerts people rather than merely blandly reporting facts.

http://gborg.postgresql.org/pipermail/slony1-commit/2005-January/000333.html
http://linuxdatabases.info/info/monitoring.html