Subject: Re: checkpoints, proper config
To: pgsql-performance@postgresql.org
References: 
 <CAEaSS0af0GpqQYSwePKRQq6RtX-x9NXDm=jOXcbcGAmGHXapFw@mail.gmail.com>
 <5669B47B.9010504@commandprompt.com>
From: Tomas Vondra <tomas.vondra@2ndquadrant.com>
Message-ID: <5669FF52.2030303@2ndquadrant.com>
Date: Thu, 10 Dec 2015 23:40:18 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.1.0
MIME-Version: 1.0
In-Reply-To: <5669B47B.9010504@commandprompt.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

On 12/10/2015 06:20 PM, Joshua D. Drake wrote:
> On 12/10/2015 01:12 AM, Tory M Blue wrote:
>
>> checkpoint_timeout = 5min
>>
>> checkpoint_completion_target = 0.9
>>
>
> The above is your problem. Make checkpoint_timeout = 1h . Also,
> considering turning synchronous_commit off.

I doubt that. The report mentioned that the checkpoints happen 0.027... 
minutes apart (assuming the minutes_between_checkpoints is computed in a 
sane way). That's way below 5 minutes, so the checkpoints have to be 
triggered by something else - probably by running out of segments, but 
we don't know the value before Tory increased it to 150.

Also, recommending synchronous_commit=off is a bit silly, because not 
only it introduces data loss issues, but it'll likely cause even more 
frequent checkpoints.

Tory, please enable logging of checkpoints (log_checkpoints=on). Also, I 
don't think it makes much sense to set

    (checkpoint_warning > checkpoint_timeout)

as it kinda defeats the whole purpose of the warning.

regards

--
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services


-- 
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance