From: | Greg Smith <gsmith(at)gregsmith(dot)com> |
---|---|
To: | Kevin Grittner <Kevin(dot)Grittner(at)wicourts(dot)gov> |
Cc: | pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: 8.3 beta testing suggestions welcome |
Date: | 2007-12-01 00:04:28 |
Message-ID: | Pine.GSO.4.64.0711301853210.15926@westnet.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On Fri, 30 Nov 2007, Kevin Grittner wrote:
>> checkpoint_segments = 50
Here's how I'm reading your data:
16:43:11 : Checkpoint start
16:44:23 : Checkpoint ends [ 1:22 long]
1:01 passes
16:45:24 : Checkpoint start
16:46:36 : checkpoint ends [1:22 long]
If you're getting a checkpoint every minute or so, the number of
checkpoint segments you're using is still far too low to really take
advantage of the new checkpoint spreading features. You want to make that
high enough that it's several minutes between them. It's not unusual to
set a much higher checkpoint_segments during bulk loading than you'd
necessarily want for actual production use.
One thing that would make your reports more useful: in the spots in your
log where you're labeling [checkpoint starting] etc., if you could include
the new message spit out by turning on checkpoint_log there it would help.
That's the one that says how much data was written.
Did you ever write something to save snapshots of pg_stat_bgwriter?
Those would be interesting to see on the same time scale as well.
--
* Greg Smith gsmith(at)gregsmith(dot)com http://www.gregsmith.com Baltimore, MD
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2007-12-01 04:37:04 | Re: PostGreSQL and recursive queries... |
Previous Message | Usama Dar | 2007-11-30 23:21:25 | Re: Release Note Changes |