Andy Lester wrote:
>
> On Sep 5, 2011, at 3:19 PM, Bruce Momjian wrote:
>
> >
> > This is an interesting bug report from March that got no replies. In my
> > testing, not only does CLUSTER not update the n_dead_tup statistics, but
> > neither does VACUUM FULL, which internally uses the CLUSTER code
> > (cluster_rel()). Is this a bug?
>
> I can't imagine how it NOT be a bug to do something that gets rid of dead tuples and then tell the user there are
deadtuples when there are actually no dead tuples.
Well, if you TRUNCATE I don't think it updates the statistics either. I
think we assume the autovacuum system will analyse the new table soon.
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +