Message-ID: <50D25A59.3060105@richardneill.org>
Date: Thu, 20 Dec 2012 00:22:49 +0000
From: Richard Neill <rn214@richardneill.org>
User-Agent: Mozilla/5.0 (X11; Linux x86_64;
	rv:16.0) Gecko/20121010 Thunderbird/16.0.1
MIME-Version: 1.0
To: Sergey Konoplev <gray.ru@gmail.com>
CC: "pgsql-performance@postgresql.org" <pgsql-performance@postgresql.org>
Subject: Re: Why does the query planner use two full indexes, when
	a dedicated partial index exists?
References: <50D22DE2.4090301@richardneill.org>
	<CAL_0b1s6_YZDdzt3NizwC8X+dYA37T-4aZOTzw3KJ3-iXuSjjw@mail.gmail.com>
	<50D2528E.7000800@richardneill.org>
	<CAL_0b1t5_h-VUE+sL2cfd1E78xYOq9T7P3n31iZhLT3vEa=mOA@mail.gmail.com>
In-Reply-To: 
 <CAL_0b1t5_h-VUE+sL2cfd1E78xYOq9T7P3n31iZhLT3vEa=mOA@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

Thanks for your help,

On 20/12/12 00:08, Sergey Konoplev wrote:
> On Wed, Dec 19, 2012 at 3:49 PM, Richard Neill <rn214@richardneill.org> wrote:
>> * The reindex solution doesn't work. I just tried it, and the query planner
>> is still using the wrong indexes.
>
> Can you show the explain analyze with tbl_tracker_performance_1_idx
> straight after reindex (eg. before it has been bloated again)?

Sure. Just done it now... the system has been fairly lightly loaded for 
the last few hours - though I did have to change the specific number of 
the parcel_id_code in the query.


fsc_log=> explain analyse select * from tbl_tracker where 
parcel_id_code=92223 and exit_state is null;

QUERY PLAN
-----------------------------------------------------------
  Index Scan using tbl_tracker_exit_state_idx on tbl_tracker 
(cost=0.00..6.34 rows=1 width=174) (actual time=0.321..1.871 rows=1 loops=1)
    Index Cond: (exit_state IS NULL)
    Filter: (parcel_id_code = 92223)
  Total runtime: 1.905 ms
(4 rows)


And now, force it, by dropping the other index (temporarily):

fsc_log=> drop index tbl_tracker_exit_state_idx;
DROP INDEX


fsc_log=> explain analyse select * from tbl_tracker where 
parcel_id_code=92223 and exit_state is null;

QUERY PLAN
---------------------------------------------------------------------
  Index Scan using tbl_tracker_performance_1_idx on tbl_tracker 
(cost=0.00..7.78 rows=1 width=174) (actual time=0.040..0.041 rows=1 loops=1)
    Index Cond: (parcel_id_code = 92223)
  Total runtime: 0.077 ms
(3 rows)


As far as I can tell, the query planner really is just getting it wrong.

BTW, there is a significant effect on speed caused by running the same 
query twice (it pulls stuff from disk into the OS disk-cache), but I've 
already accounted for this.


Richard


-- 
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance