Re: More Praise for 7.4RC2

reece(at)in-machina(dot)com (Reece Hart) writes:
> On Thu, 2003-11-13 at 10:09, scott.marlowe wrote:
>
> Do you vacuum full every so often? If not, and if you've been overflowing
> your fsm, then your tables will just grow without shrinking.
> Also, index growth could be a problem.
>
> Hmm. I didn't realize that I needed to vacuum full as well -- I
> thought vacuum was sufficient for performance gains, and that full
> reclaimed space but didn't result in significant performance
> gains. I have reindexed infrequently, but since that locks the table
> I didn't do that (or vacuum full) often. I guess I should try out
> pg_autovacuum, but I think that full vacuums only to prevent XID
> wraparound (if age>1.5B transactions), but not for compaction (is
> this correct?).

That's not quite correct. pg_autovacuum NEVER does a VACUUM FULL, and
it is not necessary to do so in order to avoid XID wraparound. A
"simple VACUUM" suffices for that purpose..

What pg_autovacuum "buys you" is mainly twofold:

1. You don't need to schedule batch jobs to vacuum things;

2. If you have heavily updated tables, it will vacuum them a lot,
which should prevent them from "blowing out" the free space map,
and allow quicker reuse of dead space.

The merits of that aren't infinite, but are not nothing, either.
--
output = reverse("ofni.smrytrebil" "@" "enworbbc")
<http://dev6.int.libertyrms.com/>
Christopher Browne
(416) 646 3304 x124 (land)