Re: Error in 9.6 documentation? checkpoint_flush_after

On Tue, Mar 5, 2019 at 09:11:14AM -0800, Andres Freund wrote:
> Hi,
>
> On 2019-03-05 11:05:31 -0500, Bruce Momjian wrote:
> > On Fri, Mar 1, 2019 at 05:58:21AM +0000, PG Doc comments form wrote:
> > > The following documentation comment has been logged on the website:
> > >
> > > Page: https://www.postgresql.org/docs/9.6/runtime-config-wal.html
> > > Description:
> > >
> > > https://www.postgresql.org/docs/9.6/runtime-config-wal.html
> > > 19.5.2
> > >
> > > checkpoint_flush_after (integer)
> > >
> > > Whenever more than checkpoint_flush_after bytes have been written while
> > > performing a checkpoint, attempt to force the OS to issue these writes to
> > > the ...
> > >
> > > bytes or pages? postgresql.conf says pages
>
> It's bytes, rounded to pages. Describing it as pages is too complicated
> because we'd need to reference the block size for it to make sense etc.
> What's the problem with describing it in bytes?

Well, if postgrsql.conf has:

checkpoint_flush_after = 2

it is intepreted in pages and output as bytes:

SHOW checkpoint_flush_after;
checkpoint_flush_after
------------------------
16kB

> > You bring up a good point. This problem affects these postgresql.conf
> > variables:
> >
> > bgwriter_flush_after
> > backend_flush_after
> > checkpoint_flush_after
> >
> > and probably more. The issue is that the value is in bytes, as opposed
> > to a value of time, like milliseconds. You can specify the bytes in
> > terms of the number of pages, e.g., 2 = 16kB, or you can specify it
> > directly in bytes, e.g., 32kB.
>
> I'm not following? Why is bytes vs time a problem? We have similar
> base-unit issues in plenty time based GUCs?

I am only pointing out that pages and literal bytes are all bytes, while
wal_writer_delay is time, e.g., 200ms.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ As you are, so once was I. As I am, so you will be. +
+ Ancient Roman grave inscription +