Re: Very slow checkpoints

Hi Ilya,

Thank you for the response.

Yes BBU is on the controller; 1024Mb. It is a HP P410i controller, with write caching turned on the controller; off on disk level.

2x 15k SAS SFF for WAL and 12x 10k SAS SFF for DB

We have tried longer settings for checkpoint_timeout, but not 1hr; so we will try that as well.

We don't at this stage have any graphs, but we will set it up over the next 24hrs at least.

Regards,
Steve

----------------------------------------
> From: ilya(dot)kosmodemiansky(at)postgresql-consulting(dot)com
> Date: Wed, 18 Mar 2015 12:42:43 +0100
> Subject: Re: [PERFORM] Very slow checkpoints
> To: steven(dot)jones1201(at)outlook(dot)com
> CC: pgsql-performance(at)postgresql(dot)org
>
> Hi Steven,
>
> On Wed, Mar 18, 2015 at 12:21 PM, Steven Jones
> <steven(dot)jones1201(at)outlook(dot)com> wrote:
>> - system is HP blade; 128Gb RAM, 2x 8-core, 12x 10k RPM RAID1+0 (database)
>
> Have you BBU on your controller? And how your controller configured, I
> mean cache mode, io mode, disk write cache mode. You have 15K SAS
> (which form factor?) under WAL and 10K SAS under database, am I
> correct?
>
>> Full iostat/iotop, configuration, checkpoint stats, etc. are pasted below
>> for completeness. Highlights are:
>> checkpoint_segments=512
>> shared_buffers=16GB
>> checkpoint_timeout=15min
>> checkpoint_completion_target=0.1
>
> It looks like your checkpoint settings are a bit strange besides of
> everything else. If you chose high value for checkpoint_segments, your
> aim is to avoid checkpoints by timeout (or vice verse). If you have
> checkpoint_segments=512, your checkpoint_timeout should be about
> 60min. And anyway - checkpoint_completion_target=0.9 or 0.7 in order
> to spread disk load between checkpoints.
>
>
>
>> 
>> sysctl settings for dirty pages
>>
>> vm.dirty_background_bytes = 0
>> vm.dirty_background_ratio = 5
>> vm.dirty_bytes = 0
>> vm.dirty_expire_centisecs = 3000
>> vm.dirty_ratio = 10
>> vm.dirty_writeback_centisecs = 500
>
> Values for this settings are really dependent of RAID (and BBU size).
>
> And about further problem description: have you any graphical
> representation of your % disc utilization?
>
> Best regards,
> Ilya
>
> --
> Ilya Kosmodemiansky,
>
> PostgreSQL-Consulting.com
> tel. +14084142500
> cell. +4915144336040
> ik(at)postgresql-consulting(dot)com
>
>
> --
> Sent via pgsql-performance mailing list (pgsql-performance(at)postgresql(dot)org)
> To make changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-performance