Chris,

> My question is whether we can get a good diagnosis of the cause of the
> situation.  It seems plausible that this might be something justifying a
> 1.0.6 release, particularly if it is a condition people could fairly
> readily fall into by accident.
>
> What was the "proximate cause"?  If we can document that, that's a good
> part of the battle all by itself.

Well, the cause is something which is an issue all by itself; client code 
opening transactions and forgetting to close them, and middleware which 
doesn't rotate out idle connections until the server hits max.  
Long-open-idle transactions are a performance problem even without Slony; 
they prevent expiring rows and make VACUUM inneffective.   BWHGATI, there 
aren't any utitilties in the PG server to specifcially kill these undead 
transactons, though one could be improvised with a simple perl script.  So 
this is mostly a Not-Slony issue.

The only issue for Slony is that, if there isn't a reason we need to be 
concerned with the servers minXID, not to hold on to s_log_1 entries until it 
goes up.

-- 
--Josh

Josh Berkus
Aglio Database Solutions
San Francisco