Message-ID: <55842859.4060102@intuvisiontech.com>
Date: Fri, 19 Jun 2015 10:34:01 -0400
From: Ian Pushee <ian@intuvisiontech.com>
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64;
 rv:31.0) Gecko/20100101 Thunderbird/31.7.0
MIME-Version: 1.0
To: pgsql-performance@postgresql.org
Subject: Slow query (planner insisting on using 'external merge' sort type)
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

Hi Folks,

This is my first time posting here, so hopefully I manage to convey all 
the information needed.
We have a simple query that just started giving us problems in 
production when the number of rows gets too large (>100k).
The issue seems to be that the planner wants to sort the rows using a 
sequential scan, rather than the index provided specifically for this 
query. This isn't a problem with low numbers of rows, but eventually the 
query outgrows work_mem and uses the disk, slowing does the query 
greatly. I know the common answer is to increase work_mem... but since 
this tables growth is unpredictable, that isn't a viable strategy.
I've tried increasing shared_buffers and effective_cache_size, but that 
doesn't appear to effect the plan chosen here. Setting 
random_page_cost=1.0 works, but I'm hoping for a more general solution 
that doesn't require setting that locally each time I run the query. I 
guess my real question is wether or not there is any way to get the 
planner to take into account the fact that it's going to need to do an 
'external merge', and that it is going to take a LONG time?

Table and Index Schemas:
CREATE TABLE events
(
   id serial NOT NULL,
   name character varying(64),
   eventspy_id integer NOT NULL,
   camera_id integer NOT NULL,
   start_time timestamp without time zone NOT NULL,
   millisecond smallint NOT NULL,
   uid smallint NOT NULL,
   update_time timestamp without time zone NOT NULL DEFAULT now(),
   length integer NOT NULL,
   objects text NOT NULL,
   priority smallint NOT NULL,
   type character varying(45) NOT NULL DEFAULT 'alarm'::character varying,
   status event_status NOT NULL DEFAULT 'new'::event_status,
   confidence smallint NOT NULL DEFAULT 100::smallint,
   CONSTRAINT events_pkey PRIMARY KEY (id)
)
WITH (
   OIDS=FALSE
);

CREATE INDEX events_confidnce
   ON events
   USING btree
   (confidence);

CREATE INDEX events_summary
   ON events
   USING btree
   (name COLLATE pg_catalog."default", eventspy_id, camera_id, type 
COLLATE pg_catalog."default", status);

Query:
SELECT name, type, eventspy_id, camera_id, status, COUNT(id), 
MAX(update_time), MIN(start_time), MAX(start_time) FROM events WHERE 
confidence>=0 GROUP BY name, eventspy_id, camera_id, type, status;

Explain Analyze outputs (links as requested):
Default plan: http://explain.depesz.com/s/ib3k
Forced index (random_page_cost=1.0): http://explain.depesz.com/s/lYaP

Software/Hardware: PGSql 9.2.1, Windows 8.1, 8GB RAM
All pgsql settings are at their defaults.

Thanks for any help you can provide,
-Ian Pushee


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