Re: new vacuum is slower for small tables

Heikki Linnakangas escribió:

> Oprofile suggests that most of the time is actually spent in
> pgstat_vacuum_stat. And more precisely in pstat_collect_oids, which is
> called by pgstat_vacuum_stat.

Hmm, that routine is expensive. Calling it for every vacuum is not good
:-( Fortunately, autovacuum calls it only once per worker run rather
than once per table. That limits the damage.

I wonder if we could do better in pgstat_vacuum_stat; for example,
scanning all of pg_proc is useless most of the time, and then it has to
fill and seq-search a hash table with all the builtins which will never
get dropped. I wonder if we could use pg_depend instead of pg_proc, and
skip pinned functions, for example.

In the end, it would be better if this function was not called at all
for user-invoked vacuum, and have autovacuum handle it. However, that
doesn't work for people who disable autovacuum.

--
Alvaro Herrera http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc.