Horton

Re: BUG: ReadStream look-ahead exhausts local buffers when effective_io_concurrency>=64

Xuneng Zhou <xunengzhou@gmail.com>

View thread View on postgr.es
From: Xuneng Zhou <xunengzhou@gmail.com>
To: Feike Steenbergen <feikesteenbergen@gmail.com>
Cc: Eduard Stepanov <crtxcz@gmail.com>, Induja Sreekanthan <indujas@google.com>, pgsql-hackers@postgresql.org, Simhachala Sasikanth Gottapu <simhachala@google.com>, Vishal Bagga <vishalbagga@google.com>, Madhukar <madhukarprasad@google.com>, Shihao Zhong <shihaozhong@google.com>, Yi Ding <yidin@google.com>, Hardik Singh Negi <hardiksnegi@google.com>
Date: 2026-07-03T03:54:31Z
Lists: pgsql-hackers 
Hi Feike,

On Wed, May 6, 2026 at 8:48 PM Feike Steenbergen
<feikesteenbergen@gmail.com> wrote:
>
>
>
> On Wed, 6 May 2026 at 14:24, Feike Steenbergen <feikesteenbergen@gmail.com> wrote:
> > For now, the problem disappears again when switching to io_method='worker'
>
> I spoke too soon it seems. It happens less frequently, but we do still have the issue even with io_method='worker'.
> Reverting to io_method='sync' as I would expect that will solve the issue.

Interestingly, I wonder why this variance would occur in production.
This is the more interesting part of this thread. I did some
investigation & experiment, pg 19 passed the reproducer even if
da6874635db is removed. A series of improvements made in April may
contribute to this. Will return soon to share the findings..









--
Regards,
Xuneng Zhou
HighGo Software Co., Ltd.