В Птн, 07.11.2003, в 16:45, Tom Lane пишет:
> Markus Bertheau <twanger@bluetwanger.de> writes:
> > It has a unique index on virtual_domain. Every day we do <10000 queries
> > of the form
> > insert into virtual_domains values (default, 'www.something.de')
>
> So you're expecting these inserts to fail because of unique index
> violations?
Yes.
>
> > The big table size slows down sequential scans significantly. What can
> > we do to prevent growth of the table?
>
> Vacuum more often. You have to get rid of the tuples that failed the
> index check. They are already in the table when the index spits up.
Does pg_autovacuum do something wrong, that it not catches this case?
--
Markus Bertheau <twanger@bluetwanger.de>