[pgpool-general: 3293] Re: High load, low CPU usage since 3.3.4

[Gersom van de Bunt]

Hi Tatsuo,

Thanks for your reply. You’re right, i noticed that wd_lifecheck_method defaults to “heartbeat” now, so i’ve also configured the heartbeat settings.

Attached the output of “show pool_status”. Unfortunately, i’m currently not able to disable watchdog without interruption. However, i could try to reproduce this behaviour on a test setup, and turn of watchdog there.

Regards,

Gersom



-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: pgpool-show_pool_status.txt
URL: <http://www.sraoss.jp/pipermail/pgpool-general/attachments/20141118/90ca58c8/attachment-0002.txt>
-------------- next part --------------



On 18 Nov 2014, at 01:49, Tatsuo Ishii <ishii at postgresql.org> wrote:

> pgpool-II changes watchdog functionality between 3.2 and 3.3. I don't
> what is your actual configuration (you only showed old configuration
> contents, you should use "show pool_status" to display actual
> settings) but I suspect this is because of the changes in
> watchdog. Can you try to turn off watchdog and how the load average
> changes?
> 
> Best regards,
> --
> Tatsuo Ishii
> SRA OSS, Inc. Japan
> English: http://www.sraoss.co.jp/index_en.php
> Japanese:http://www.sraoss.co.jp
> 
>> Hi all,
>> 
>> Recently upgraded a production system from 3.2.4-2.1 to 3.3.4-1 (both backported Debian packages). With 3.2.4 the cpu usage was ~10% (8 cores), and load ~0.1. After upgrade, without any changes to configuration, cpu usage is the same, however, system load is now about 10.0 (5min. average). Only similar report i could find is from 2012:
>> 
>> http://www.sraoss.jp/pipermail/pgpool-general/2012-July/000827.html
>> 
>> This suggests the issue was resolved in 3.1-stable, so this is probably another issue we’re facing. Some more info;
>> 
>> - PGpool runs on dedicated hosts, in failover (watchdog) setup
>> - PGpool nodes have been upgrade from Debian squeeze to wheezy, while upgrading from 3.2.3 to 3.3.4
>> - No changes (configuration and/or updates) on backend (postgresql) nodes
>> - The load does not seem to affect functionality in any way (not in the application and neither in pcp commands or responsiveness)
>> - No changes in (use of) system resources are noticeable; no processes in IOwait, not more (or less) connections / changes in state of connections to/from pgpool)
>> 
>> I’ve attached our PGpool settings. Any ideas?
>> 
>> Thanks!
>> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 496 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://www.sraoss.jp/pipermail/pgpool-general/attachments/20141118/90ca58c8/attachment-0002.sig>