Re: RT3.4 query needed a lot more tuning with 9.2 than it did with 8.1

From: Christoph Berg <cb(at)df7cb(dot)de>
To: pgsql-performance(at)postgresql(dot)org
Subject: Re: RT3.4 query needed a lot more tuning with 9.2 than it did with 8.1
Date: 2013-05-15 06:48:20
Message-ID: 20130515064819.GD704@msgid.df7cb.de
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Re: Tom Lane 2013-05-06 <1583(dot)1367858328(at)sss(dot)pgh(dot)pa(dot)us>
> The newer rowcount estimates are much further away from reality:
>
> > Unique (cost=1117.67..1118.46 rows=9 width=1115) (actual time=82.646..85.695 rows=439 loops=1)
>
> > Unique (cost=784205.94..796940.08 rows=145533 width=1061) (actual time=9710.683..9713.175 rows=439 loops=1)
>
> Has the new DB been analyzed? Maybe you had custom stats targets in
> the old DB that didn't get copied to the new one?

The new DB was analyzed with various stats targets, including values
that were higher that anything we would have used in 8.1. I don't
think we had per-table settings in there (the actual DB is now gone
for good).

Christoph
--
cb(at)df7cb(dot)de | http://www.df7cb.de/

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Christoph Berg 2013-05-15 06:52:29 Re: RT3.4 query needed a lot more tuning with 9.2 than it did with 8.1
Previous Message Robert Haas 2013-05-14 12:44:56 Re: Lock and pg_stat