Re: bgwriter, checkpoints, curious (seeing delays) - Mailing list pgsql-performance

From Tory M Blue
Subject Re: bgwriter, checkpoints, curious (seeing delays)
Date
Msg-id 8a547c841002261101l6bf9bd37j8b2068a0190bedff@mail.gmail.com
Whole thread Raw
In response to Re: bgwriter, checkpoints, curious (seeing delays)  (Devrim GÜNDÜZ <devrim@gunduz.org>)
List pgsql-performance
2010/2/25 Devrim GÜNDÜZ <devrim@gunduz.org>:
> On Thu, 2010-02-25 at 23:01 -0800, Tory M Blue wrote:
>
>> Checkpoint_timeout is the default and that looks like 5 mins (300
>> seconds). And is obviously why I have such a discrepancy between time
>> reached and requested.
>
> If you have a high load, you may want to start tuning with 15 minutes,
> and bump it to 30 mins if needed. Also you may want to decrease segments
> value based on your findings, since increasing only one of them won't
> help you a lot.
>
> As I wrote before, pg_stat_bgwriter is your friend here.

Actually these servers have almost no load. Really they run very cool

Load Average (for the day):
Cur: 0.16 Load
Avg: 0.22 Load

So I don't think it's load  No network issues (that I've found)  and
while the server will eventually eat all the memory it's currently
sitting with 4 gb free.

Mem Total (for the day):
Cur: 25.55 GBytes
Avg: 25.55 GBytes
Max: 25.55 GBytes
    Mem Available (for the day):
Cur: 4.72 GBytes
Avg: 5.15 GBytes
Max: 5.71 GBytes
    Bytes Used (for the day):
Cur: 20.83 GBytes
Avg: 20.40 GBytes
Max: 21.20 GBytes


Thanks for your pointers, I'm continuing to look and will do some
tests today. I also hear you about fsync and will do some testing here
to see why this was set (been running for 4-6 years), and that setting
was probably set way way back in the day and it's survived each
upgrade/hardware/storage update.

Tory

pgsql-performance by date:

Previous
From: Tory M Blue
Date:
Subject: Re: bgwriter, checkpoints, curious (seeing delays)
Next
From: "Jorge Montero"
Date:
Subject: Re: bgwriter, checkpoints, curious (seeing delays)