Re: Forcing use of specific index - Mailing list pgsql-performance

From Junaili Lie
Subject Re: Forcing use of specific index
Date
Msg-id 8d04ce9905060318523d368995@mail.gmail.com
Whole thread Raw
In response to Forcing use of specific index  (Tobias Brox <tobias@nordicbet.com>)
List pgsql-performance
HI all,
I also would like to know if there is a way to force a use of a
specific index for a specific query. I am currently using Postgresql
7.4.6

In my case I have a relatively big table (several millions of records)
that are frequently used to join with other tables (explicit join or
through view).
The table has several indices, some are single column and some are multi column.
Some queries are faster if using single colum index while other are
faster using multi column indexes.
I have play around with SET STATISTICS, but it doesn't seem to make
any differences (I tried to set it to 1000 one time, but still the
same). I did analyze and vacuum after SET STATISTICS.
Any pointer on how to do this is greatly appreciated.
Thank you in advance,


J



On 6/1/05, Tobias Brox <tobias@nordicbet.com> wrote:
> Is it any way to attempt to force the planner to use some specific index
> while creating the plan?  Other than eventually dropping all the other
> indices (which is obiously not a solution in production setting anyway)?
>
> I have one case where I have added 16 indices to a table, many of them
> beeing partial indices.  The table itself has only 50k of rows, but are
> frequently used in heavy joins.  I imagine there can be exponential order on
> the number of alternative paths the planner must examinate as function of
> the number of indices?
>
> It seems to me that the planner is quite often not choosing the "best"
> index, so I wonder if there is any easy way for me to check out what the
> planner think about a specific index :-)
>
> --
> Tobias Brox, Beijing
>
> ---------------------------(end of broadcast)---------------------------
> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org
>

pgsql-performance by date:

Previous
From: Christopher Browne
Date:
Subject: Re: Insert slow down on empty database
Next
From: hubert lubaczewski
Date:
Subject: strategies for optimizing read on rather large tables