Re: Autovacuum Improvements

Matthew T. O'Connor <matthew@zeut.net>

From: Matthew O'Connor <matthew@zeut.net>
To: Glen Parker <glenebob@nwlink.com>
Cc: Postgres general mailing list <pgsql-general@postgresql.org>
Date: 2006-12-20T01:33:59Z
Lists: pgsql-hackers
Glen Parker wrote:
>> I wonder if the simple solution is to just have a cron script modify
>> postgresql.conf and pg_ctl reload.  That seems very flexible, or have
>> two postgresql.conf files and move them into place via cron.
> 
> I'd still prefer to vacuum on demand actually.  Rather than hope that 
> autovacuum hit all the nastiest tables, I'd like to be able to record 
> the fact that tables (x,y,z) were vacuumed and how long it took.  I want 
> the logic autovacuum uses to determine if a table needs vacuuming, but 
> I'd rather do the actual vacuuming myself.
> 
> I'd also like to use some of this information to issue reindex and 
> cluster commands only when they're needed.  In fact, on days when I 
> cluster, there's no need whatsoever to also vacuum those tables.  This 
> is something that autovacuum won't do at all.
> 
> If the best I got was access to the same information autovacuum uses to 
> make its decisions, I'd be pretty happy

Well take a look at the autovacuum code, also, before autovacuum was 
integrated into core it existed as a libpq based contrib application so 
you can look in one of the older branches for that code.