Re: Re: dramatic slowdown. . .fixed by vacuum

On Fri, 21 Jul 2000, Joe Slag wrote:

> On Fri, Jul 21, 2000 at 12:36:35PM -0700, WOLF, PATRICK wrote:
> > Try running vacuum on the table or the database. Here's an excerpt from the
> > man on vacuum:
> >
> [snip]
> >
> >
>
> Thanks to all who responded. I vacuumed out foo, and sure enough
> the select time is down to 10 seconds again.
>
> I see in the docs the suggestion:
>
> We recommend that active production databases be VACUUMM-ed nightly
>
> Is this how people tend to do their vacuuming? Does anyone do programmatic
> vacuums instead of / in addition to a nightly run? Is vacuuming mainly
> necessary after big deletes, or are there other common situations
> requiring it?

UPDATEs are a combination of 'INSERT new tuple/mark old as DELETEd', so
for each UPDATE, you are adding one more tuple to the table but not
removing anything. VACUUM removes that DELETEd tuple.