Subject: Re: Query planning on partitioned table causes postgres 13.4 to
 consume all memory
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-bugs@lists.postgresql.org
References: <4a3115d4-0fb2-e214-93e3-9a9d0974b883@deepbluecap.com>
 <1556207.1632067384@sss.pgh.pa.us>
From: Duncan Sands <duncan.sands@deepbluecap.com>
Message-ID: <98598292-22bb-1304-656f-773f3677a0a8@deepbluecap.com>
Date: Mon, 20 Sep 2021 14:10:17 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.13.0
MIME-Version: 1.0
In-Reply-To: <1556207.1632067384@sss.pgh.pa.us>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Archived-At: 
 <https://www.postgresql.org/message-id/98598292-22bb-1304-656f-773f3677a0a8%40deepbluecap.com>
Precedence: bulk

Hi Tom,

On 19/09/2021 18:03, Tom Lane wrote:
> Duncan Sands <duncan.sands@deepbluecap.com> writes:
>> [ planning DELETE on a thousand-partition table takes forever ]
> 
> FWIW, this situation has been very much improved for v14 [1].

thanks, part (2) of that commit indeed looks like it should solve it.

Best wishes, Duncan.

> In older branches, the best advice I can give you is "don't use
> so many partitions".  Especially not with hash partitioning,
> where the query WHERE clause generally won't translate to any
> useful pruning of the partitions.
> 
> (Personally, I think that hash partitioning is an evil that
> we shouldn't have implemented at all.  Or at least there
> should be stronger warnings about it in the manual than there
> are now.)
> 
> 			regards, tom lane
> 
> [1] https://git.postgresql.org/gitweb/?p=postgresql.git&a=commitdiff&h=86dc90056
>