On Thu, May 30, 2013 at 12:49 AM, Stephen Scheck
<singularsyntax(at)gmail(dot)com> wrote:
> If this hypothesis is correct, doing a vacuum should free up dead pages and
> your size expectations should be more accurate. And if that's the case
> putting more intelligence into the application could mitigate some of the
> update growth (predicting what page temporally similar updates will go to
> and grouping them into a single transaction, for instance).
>
Seems correct to me, according to this
http://www.postgresql.org/docs/current/static/lo.html
I would give a try to vacuumlo to see if the size decreases, in such
case that is the right explaination.
Luca