Checkpoints writes

Started by Cyril SCETBONover 17 years ago3 messagesgeneral
Jump to latest
#1Cyril SCETBON
scetbon@echo.fr

Hi guys,

What's the way to count the read/write bytes of the checkpoint process
before 8.3 (no pg_stat_bgwriter view :-[ )
I want to distinguish bytes written by checkpoints and others written by
the background process

thanks
--
Cyril SCETBON

#2Greg Smith
gsmith@gregsmith.com
In reply to: Cyril SCETBON (#1)
Re: Checkpoints writes

On Thu, 7 Aug 2008, Cyril SCETBON wrote:

What's the way to count the read/write bytes of the checkpoint process
before 8.3 (no pg_stat_bgwriter view :-[ ) I want to distinguish bytes
written by checkpoints and others written by the background process

The reason that view was added was because it's really hard to figure that
out in earlier versions. Theoretically you could have some operating
system level program that tracked I/O on a per-process basis, noting which
one was the background writer process and counting those separately. I
found it easier to work on adding the counters instead.

It's not really complete, but I did have a functional prototype of a
pg_stat_bgwriter implementation that worked against 8.2 if that helps you
any: http://www.westnet.com/~gsmith/content/postgresql/perfmon82.htm

--
* Greg Smith gsmith@gregsmith.com http://www.gregsmith.com Baltimore, MD

#3Cyril SCETBON
scetbon@echo.fr
In reply to: Greg Smith (#2)
Re: Checkpoints writes

Greg Smith wrote:

On Thu, 7 Aug 2008, Cyril SCETBON wrote:

What's the way to count the read/write bytes of the checkpoint
process before 8.3 (no pg_stat_bgwriter view :-[ ) I want to
distinguish bytes written by checkpoints and others written by the
background process

The reason that view was added was because it's really hard to figure
that out in earlier versions. Theoretically you could have some
operating system level program that tracked I/O on a per-process
basis, noting which one was the background writer process and counting
those separately. I found it easier to work on adding the counters
instead.

It's not really complete, but I did have a functional prototype of a
pg_stat_bgwriter implementation that worked against 8.2 if that helps
you any: http://www.westnet.com/~gsmith/content/postgresql/perfmon82.htm

a really great job greg. I'll test it.

Thanks

--
* Greg Smith gsmith@gregsmith.com http://www.gregsmith.com Baltimore, MD

--
Cyril SCETBON