[Pgpool-general] insert_lock leaks

[Simone Tregnago]

Yoshiyuki Asaba wrote:
> Hi,
> 
> From: Simone Tregnago <simonetregnago at grivaonline.com>
> Subject: Re: [Pgpool-general] insert_lock leaks
> Date: Thu, 29 May 2008 11:19:02 +0200
> 
> 
> Could you provide a reproducible program?
> 

That's very hard to do since the leaks comes using our ERP with hundreds 
of opened connections ( in order to work properly I had to set 
num_init_children=1000 and max_pool=4). I can't imagine how to reproduce 
a test software: until now I wasn't able to guess the guilty context.
Maybe, if you have suggestions on what to look ( or to log ) in order to 
help me to identify something...
A first hypothesis is that the leaks remains with tables with serial 
fields, but I can't be sure of that.
Yesterday I killed 3-4 zombie connections that took one table locked.
Note that the ERP don't use PostgreSQL locks, so only PgPool makes 
explicit locks.

Thanks for you attention,
Simone Tregnago