could that because of my system is really busy?
1, postgresql always have 400+ connections(dozens of python process using client pool)
2, the query peak is 50+/s
3, I do have some bad performance sql executing periodically, need 100+ second to complete. could those bad performance sql influence others? because when I execute those simple sql directly, they was fast. but the slow query log shows it took too much time.
On Wed, Jul 11, 2012 at 4:23 PM, Albe Laurenz
<laurenz.albe@wien.gv.at> wrote:
Yan Chunlu wrote:
> I have logged one day data and found the checkpoint is rather
frequently(detail:
>
https://gist.github.com/3088338). Not sure if it is normal, but the
average time of checkpoint is
> about 100sec~200sec, it seems related with my settings:
>
> 574 checkpoint_segments = 64
> 575 wal_keep_segments = 5000
>
> I set checkpoint_segments as a very large value which is because
otherwise the slave server always can
> not follow the master, should I lower that value?
wal_keep_segments has no impact on checkpoint frequency and intensity.
You are right that your checkpoint frequency is high. What is your value
of checkpoint_timeout?
You can increase the value of checkpoint_segments to decrease the
checkpoint frequence, but recovery will take longer then.
> or the slow query is about something else? thanks!
I guess the question is how saturated the I/O system is during
checkpoints. But even if it is very busy, I find it hard to believe
that such a trivial statement can take extremely long.
Yours,
Laurenz Albe