Message-ID: <e79d93fa-44b6-49d6-9761-5cffa59930c6@gmail.com>
Date: Thu, 3 Apr 2025 08:32:44 +0200
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: partition table optimizer join cost misestimation
To: James Pang <jamespang886@gmail.com>,
 pgsql-performance@lists.postgresql.org
References: 
 <CAHgTRfedznOOrDxLhvDCHYhTMDvsbfE4uWCmxBPywcOS-GikXg@mail.gmail.com>
 <a05e0fed-dc66-4873-ac8c-4a8352c4e69d@gmail.com>
 <CAHgTRfdOsgSkHcgB+q-=M-+hb_jyyV1icwTSaCHFbby8C-CUxQ@mail.gmail.com>
Content-Language: en-AU
From: Andrei Lepikhov <lepihov@gmail.com>
In-Reply-To: 
 <CAHgTRfdOsgSkHcgB+q-=M-+hb_jyyV1icwTSaCHFbby8C-CUxQ@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Archived-At: 
 <https://www.postgresql.org/message-id/e79d93fa-44b6-49d6-9761-5cffa59930c6%40gmail.com>
Precedence: bulk

On 4/3/25 02:46, James Pang wrote:
> Andrei,
>     Yes, from explain output, since optimizer already get the 
> merge_append cost but not take account into total cost, that make a big 
> difference.  I shared table DDLs and explain analyze,buffers output , I 
> think  the data maybe generated by other way to reproduce this issue. 
> sorry for not sharing the commercial production data here.
I think it is almost enough to identify the issue. Let me ask you the 
following questions:
1. Can you provide min and max values in columns cc.sss_id and aa.mmm_id?
2. How often do you analyze your *parent* tables tablea and tablec? 
Remember, if you want to build statistics on a partitioned table (not a 
partition), you have to explicitly call

ANALYZE tablea,tablec;

mentioning these tables in the analyze list.

3. May you provide a dump of pg_statistic on attributes cc.sss_id and 
aa.mmm_id?
4. Is there a possibility of changing a single code line and rebuilding 
your DB instance to check a conjecture?

-- 
regards, Andrei Lepikhov