Re: Slow query after upgrade from 9.0 to 9.2

From: Jeff Janes <jeff(dot)janes(at)gmail(dot)com>
To: Andrzej Zawadzki <zawadaa(at)wp(dot)pl>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: Slow query after upgrade from 9.0 to 9.2
Date: 2013-01-11 17:12:16
Message-ID: CAMkU=1z0iabTNxVHKRbW5oKW+gcCac+O9x4OpETa_ygRdqomkA@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On Fri, Jan 11, 2013 at 12:13 AM, Andrzej Zawadzki <zawadaa(at)wp(dot)pl> wrote:
> On 10.01.2013 19:17, Jeff Janes wrote:

>> Also, if you temporarily set enable_seqscan=off on 9.2, what plan do
>> you then get?
>
> Plan is different.
>

> " Join Filter: (sa.arrear_import_id = ai.id)"

It is hard to imagine why it is not using
sygma_arrear_arrear_import_id_idx for this given the plan is now
accessing the index anyway. Have the types or encodings or collations
somehow become incompatible so that this index can no longer fulfill
it?

What if you just write a very simple join between the two tables with
the above join condition, with another highly selective condition on
ai?

Cheers,

Jeff

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Horst Dehmer 2013-01-11 23:55:36 Insert performance for large transaction with multiple COPY FROM
Previous Message Matheus de Oliveira 2013-01-11 11:55:07 Re: Partition insert trigger using C language