On Wed, Apr 20, 2022 at 7:50 PM Thomas Munro <thomas.mu...@gmail.com> wrote:
> As for your general question, I think you must be right. From a quick
> rummage around in the commit log, it does appear that commit cddca5ec
> (2009), which introduced pgBufferUsage, always bumped the counters
> unconditionally. It predated track_io_timing by years (40b9b957694
> (2012)), and long before that the Berkeley code already had a simpler
> thing along those lines (ReadBufferCount, BufferHitCount etc). I
> didn't look up the discussion, but I wonder if the reason commit
> 9d3b5024435 (2011) introduced VacuumPage{Hit,Miss,Dirty} instead of
> measuring level changes in pgBufferUsage is that pgBufferUsage didn't
> have a dirty count until commit 2254367435f (2012), and once the
> authors had decided they'd need a new special counter for that, they
> continued down that path and added the others too?
I knew about pgBufferUsage, and I knew about
VacuumPage{Hit,Miss,Dirty} for a long time. But somehow I didn't make
the very obvious connection between the two until today. I am probably
not the only one.
--
Peter Geoghegan
