> We discovered a problem this morning – a field had been added to the master
> database but not the slave database. Slony ceased replicating
> (understandably) but offered no notification. We discovered this about 12
> hours later when people started reporting problems.

In case you aren't aware of this, you should have done a SLONIK
EXECUTE SCRIPT to add a new column to a replicated table. DDL triggers
are sort of on the agenda for Postgres 9.1 - Jan Wieck has expressed
interest in implementing them - so there may come a day when what you
did actually does work, in a future version of Slony.

Slony explicitly disclaims responsibility for determining when
replication is broken in the case of failover. The docs say "Slony-I
does not provide any automatic detection for failed systems", and I
think that's applicable for your situation too. That said, it may be
possible to write a pl/perlu trigger on one of the tables in the
_your_replication_set schema, that sends notification when an event
backlog reaches some arbitrary threshold value. Perhaps one of the
Slony hackers can weigh in on this. However, this is just as ad-hoc as
what you're already doing, and I'm not sure it's an improvement at
all.

Regards,
Peter Geoghegan