Re: Slow query: select * order by XXX desc offset 10 limit 10 - Mailing list pgsql-general

On 14 Oct 2011, at 11:14, Alexander Farber wrote:

> I've added 3 new indices on both tables:
>
>
> quincy=> \d quincynoreset
>               Table "public.quincynoreset"
>   Column    |            Type             |   Modifiers
> -------------+-----------------------------+---------------
> appsversion | character varying(30)       |
> beta_prog   | character varying(20)       |
> category    | character varying(120)      |
> catinfo     | character varying(120)      |
> details     | character varying(50)       |
> devinfo     | character varying(4000)     |
> emailid     | character varying(16)       |
> email       | character varying(320)      |
> formfactor  | character varying(10)       |
> id          | character varying(20)       | not null
> imei        | character varying(25)       |
> name        | character varying(20)       |
> osversion   | character varying(30)       |
> pin         | character varying(12)       |
> qdatetime   | timestamp without time zone |
> copied      | timestamp without time zone | default now()
> Indexes:
>    "quincynoreset_pkey" PRIMARY KEY, btree (id)
>    "quincynoreset_appsversion_index" btree (appsversion)
>    "quincynoreset_osversion_index" btree (osversion)
>    "quincynoreset_qdatetime_index" btree (qdatetime)

(...)

> quincy=> explain select to_char(qdatetime, 'YYYY-MM-DD') as
> QDATETIME_2,ID,NAME,CATEGORY,APPSVERSION,OSVERSION,DETAILS,DEVINFO
> from quincyview where qdatetime <= now() order by QDATETIME desc
> offset 10 limit 10;
>
>
>      QUERY PLAN
>
>
>
----------------------------------------------------------------------------------------------------------------------
>
----------------------------------------------------------------------------------------------------------------------
>
----------------------------------------------------------------------------------------------------------------------
> ----------------------------------------------
> Limit  (cost=558551.88..558551.91 rows=10 width=1172)
>   ->  Sort  (cost=558551.86..560883.79 rows=932773 width=1172)
>         Sort Key: quincyview.qdatetime
>         ->  Subquery Scan quincyview  (cost=482428.59..533731.10
> rows=932773 width=1172)
>               ->  Unique  (cost=482428.59..522071.44 rows=932773 width=252)
>                     ->  Sort  (cost=482428.59..484760.52 rows=932773 width=252)
>                           Sort Key: quincynoreset.qdatetime,
> quincynoreset.id, quincynoreset.name, quincynoreset.cate
> gory, quincynoreset.appsversion, quincynoreset.osversion,
> quincynoreset.beta_prog, quincynoreset.catinfo, quincynorese
> t.details, quincynoreset.devinfo, quincynoreset.email,
> quincynoreset.emailid, quincynoreset.imei, quincynoreset.pin, q
> uincynoreset.formfactor, quincynoreset.copied
>                           ->  Append  (cost=0.00..55177.71
> rows=932773 width=252)
>                                 ->  Seq Scan on quincynoreset
> (cost=0.00..39171.89 rows=807446 width=242)
>                                       Filter: (qdatetime <= now())
>                                 ->  Seq Scan on quincytrack
> (cost=0.00..6678.09 rows=125327 width=315)
>                                       Filter: (qdatetime <= now())
> (12 rows)


An explain analyse would have been a bit more informative.

Anyway, I think you get the sequential scans because the UNION requires to sort all the data from both tables to
guaranteethat the results are unique (hence that long Sort Key at the 7th line of explain output). 
For that reason, an index on qdatetime alone won't help much, especially when most of your data has qdatetime <= now(),
whichis probably the case. 

It doesn't matter that you only want 10 results from that set, the database will first have to figure out which those
rowsare. That gets more complicated because they can come from two different tables, due to the UNION. 


Do you really need unique results from that view, or are duplicates acceptable (one from each table)? In that case, try
UNIONALL instead of UNION. 

If you do need unique results, then you could create an index on the combination of all those fields. That should take
outthe need for those sequential scans. 

Alban Hertroys

--
The scale of a problem often equals the size of an ego.