On Fri, May 10, 2024 at 7:26 PM Nazir Bilal Yavuz <byavu...@gmail.com> wrote:
>
> Hi,
>
> Thank you for working on this!
>
> On Wed, 1 May 2024 at 06:37, Masahiko Sawada <sawada.m...@gmail.com> wrote:
> >
> > Thank you for further testing! I've pushed the patch.
>
> I realized a behaviour change while looking at 'Use pgBufferUsage for
> block reporting in analyze' thread [1]. Since that change applies here
> as well, I thought it is better to mention it here.
>
> Before this commit, VacuumPageMiss did not count the blocks if its
> read was already completed by other backends [2]. Now,
> 'pgBufferUsage.local_blks_read + pgBufferUsage.shared_blks_read'
> counts every block attempted to be read; possibly double counting if
> someone else has already completed the read.
True. IIUC there is such a difference only in HEAD but not in v15 and
v16. The following comment in WaitReadBuffers() says that it's a
traditional behavior that we count blocks as read even if someone else
has already completed its I/O:
/*
* We count all these blocks as read by this backend. This is traditional
* behavior, but might turn out to be not true if we find that someone
* else has beaten us and completed the read of some of these blocks. In
* that case the system globally double-counts, but we traditionally don't
* count this as a "hit", and we don't have a separate counter for "miss,
* but another backend completed the read".
*/
So I think using pgBufferUsage for (parallel) vacuum is a legitimate
usage and makes it more consistent with other parallel operations.
Regards,
--
Masahiko Sawada
Amazon Web Services: https://aws.amazon.com
