postgresql 10.1 wrong plan in when using partitions bug

From: Mariel Cherkassky <mariel(dot)cherkassky(at)gmail(dot)com>
To: PostgreSQL mailing lists <pgsql-performance(at)postgresql(dot)org>
Subject: postgresql 10.1 wrong plan in when using partitions bug
Date: 2018-02-04 10:14:04
Message-ID: CA+t6e1m54jrEj7spf9ArH5BJiO-Qznn=_YCRGWWMtohvXg7mEg@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Hi,
I configured range partitions on a date column of my main table(log_full).
Each partition represents a day in the month. Every day partition has a
list parition of 4 tables on a text column.

log_full
log_full_01_11_2017 -->
log_full _01_11_2017_x1
log_full _01_11_2017_x2
log_full _01_11_2017_x3
log_full _01_11_2017_x4
log_full_02_11_2017
log_full _02_11_2017_x1
log_full _02_11_2017_x2
log_full _02_11_2017_x3
log_full _02_11_2017_x4

and so on....

The date column consist of date in the next format : YYYY-MM-DD HH:24:SS
for example : 2017-11-01 00:01:40

I wanted to check the plan that I'm getting for a query that is using the
date column and it seems that the planner choose to do seq scans on all
tables.

-Each partition consist from 15M rows.
I have about 120 partitions.

The query :
explain select count(*) from log_full where end_date between
to_date('2017/12/03','YY/MM/DD') and to_date('2017/12/03','YY/MM/DD');

The output is too long but it do full scans on all paritions...
any idea what can be the problem? Is it connected to the date format ?

Thanks , Mariel.

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Tomas Vondra 2018-02-04 12:03:27 Re: postgresql 10.1 wrong plan in when using partitions bug
Previous Message Mariel Cherkassky 2018-02-04 09:28:47 Re: postgresql 10.1 scanning all partitions instead of 1