Re: Excessive growth of pg_attribute and other system tables

After takin a swig o' Arrakan spice grog, scrawford@pinpointresearch.com (Steve Crawford) belched out:
> On Thursday 17 March 2005 3:51 pm, Tom Lane wrote:
>> Steve Crawford <scrawford@pinpointresearch.com> writes:
>> > My autovacuum config is running and I do see regular periodic
>> > vacuums of these pg_ tables but still they grow.
>>
>> Do you have the FSM settings set large enough to account for all
>> the free space?
>
> max_fsm_pages = 20000
> max_fsm_relations = 1000

20000 is definitely way too low.  It's not enough to track the dead
pages in pg_attribute alone, which looks to have the better part of
80K dead pages.

I'd increase that to about 200000, straight off.

It seems curious that you have so many tuples getting killed off in
this table; are you generating a lot of temp tables continually?

By the way, you should be vacuuming pg_attribute _way_ more often, as
it shouldn't have gotten as big if you did so...

Doing a reindex and doing (once!) a VACUUM FULL should help bring the
size down; vacuuming the table more often should keep size down...
--
output = reverse("moc.liamg" "@" "enworbbc")
http://cbbrowne.com/info/postgresql.html
"To do is to be."  -- Aristotle
"To be is to do."  -- Socrates
"Do be do be do."  -- Sinatra
"Do be a do bee."  -- Miss Sally of Romper Room fame.
"Yabba dabba do."  -- Fred Flintstone
"DO...BEGIN..END"  -- Niklaus Wirth