Re: heavily contended lwlocks with long wait queues scale badly

From: Nathan Bossart <nathandbossart(at)gmail(dot)com>
To: "Jonathan S(dot) Katz" <jkatz(at)postgresql(dot)org>
Cc: Andres Freund <andres(at)anarazel(dot)de>, Bharath Rupireddy <bharath(dot)rupireddyforpostgres(at)gmail(dot)com>, pgsql-hackers(at)postgresql(dot)org, Thomas Munro <thomas(dot)munro(at)gmail(dot)com>, Yura Sokolov <y(dot)sokolov(at)postgrespro(dot)ru>
Subject: Re: heavily contended lwlocks with long wait queues scale badly
Date: 2024-01-11 03:17:47
Message-ID: 20240111031747.GA3376512@nathanxps13
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Mon, Nov 21, 2022 at 10:31:14AM -0500, Jonathan S. Katz wrote:
> On 11/20/22 2:56 PM, Andres Freund wrote:
>> I still think it might be worth to backpatch in a bit, but so far the votes on
>> that weren't clear enough on that to feel comfortable.
>
> My general feeling is "yes" on backpatching, particularly if this is a bug
> and it's fixable without ABI breaks.

Now that commit a4adc31 has had some time to bake and concerns about
unintended consequences may have abated, I wanted to revive this
back-patching discussion. I see a few possibly-related reports [0] [1]
[2], and I'm now seeing this in the field, too. While it is debatable
whether this is a bug, it's a quite nasty issue for users, and it's both
difficult to detect and difficult to work around.

Thoughts?

[0] https://postgr.es/m/CAM527d-uDn5osa6QPKxHAC6srOfBH3M8iXUM%3DewqHV6n%3Dw1u8Q%40mail.gmail.com
[1] https://postgr.es/m/VI1PR05MB620666631A41186ACC3FC91ACFC70%40VI1PR05MB6206.eurprd05.prod.outlook.com
[2] https://postgr.es/m/dd0e070809430a31f7ddd8483fbcce59%40mail.gmail.com

--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Andy Fan 2024-01-11 03:17:52 Re: the s_lock_stuck on perform_spin_delay
Previous Message jian he 2024-01-11 03:13:35 Re: POC PATCH: copy from ... exceptions to: (was Re: VLDB Features)