[pgpool-general: 2879] Re: Problems with the LOCK patch in Bind phase included in 3.3.2
nagata at sraoss.co.jp
Mon May 26 17:59:50 JST 2014
On Mon, 19 May 2014 12:20:35 +0200
Jose Maria Alvarez Fernandez <josemaria.alvarez at 2020mobile.es> wrote:
> Hi all,
> I am writing because I would need some help with a strange behaviour we are
> having with our pgpool2 installation.
> Our deployment includes two Postgresql 9.3.4 servers in synchronous
> streaming replication, and a pgpool2 cluster with watchdog configured to
> access them, in replication mode and with load balancing.
> When we deployed the 3.3.2 versión, we saw an unexpected behaviour that
> wasn't there in the 3.3.1 release. Our application started to get deadlocks
> when introducing data from different threads simultaneously. I think I've
> narrowed the problem to this bugfix introduced in 3.3.2:
> Fix data inconsistency problem with native replication mode + extended
> protocol case
> Per bug report by Steve Kuekes in [pgpool-general: 2142].
How did you check the commit is the problem? By usign git bisect?
For analysis, the exact log messages of pgpool with log_statement=on
and backend log are helpful for me. Could you please send me this?
> It seems that, as our application is using OpenJPA to manage data, and
> it doesn't ensure the insert order, some of the inserts trying to get
> a lock in a table that other thread as already locked, but this thread
> is also trying to acquire a lock in a table that the other thread has
> already locked, due to the unordered insert statements.
> Does anyone have an idea of how can we workaround this, to keep pgpool2
> versions up to date? I haven't found an option to force OpenJPA to "order"
> the inserts.
> Thank you all,
> Jose Maria.
> Brightstar 20:20 mobile
> C/ Arroyo de los Prados. 5. P.I. Las Arenas. 28320 (MADRID) Spain
> This e-mail has been sent by Brightstar 20:20 mobile.
> Telephone number: +34 911 032 160.
> Please see our website for further information, www.brightstar-2020.es/<http://www.2020mobile.com/confidentiality-notice/>
> *P please consider the environment and only print this if required*
Yugo Nagata <nagata at sraoss.co.jp>
More information about the pgpool-general