Jack Orenstein wrote:
> 1) Why does the tiny table bloat during a vacuum? Is it because the
> scan of the huge table is run as a transaction, forcing maintenance of
> dead versions of the tiny table's one row?
Yes.
> 2) Why does the bloat resolve itself? We're not doing any full
> vacuums.
Probably the one live tuple bounces to the first page at some point and
then the rest of the pages are truncated by vacuum.
> We're in the process of upgrading to 8.3.4, so I'd appreciate any
> throughs on whether and how this behavior will change with the newer
> release.
In 8.3, vacuuming the big table will not delay dead tuple removal of the
small table.
--
Alvaro Herrera http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc.