On Wed, 17 Apr 2002, Dima Tkach wrote:
> Tom Lane wrote:
>
> >
> >I wouldn't recommend a VACUUM FULL at all. Just do plain VACUUMs on
> >a regular basis, and accept the 10% or so storage overhead.
> >
> >VACUUM FULL is good for the sort of situation where you've updated all
> >or most of the rows at one time, and now you have a factor-of-2 storage
> >overhead; you need to physically compact the table. But the price of
> >doing that is high enough that I wouldn't do it to save 10-15%.
> >
> > regards, tom lane
> >
> I am not worried about storage overhead at all at this point, but rather
> about performance degradation when it
> has to scan through all those dead tuples in the table and there are
> LOTS of them :-(
In the 10% case, you should be within the realm where the table's steady
state size is around that much more with reasonable frequency normal
VACUUMs and an appropriately sized free space map.