[Pgpool-general] pgpool dropping db node

[Yoshiyuki Asaba]

Hi,

From: "Bruno Lustosa" <bruno.lists at gmail.com>
Subject: [Pgpool-general] pgpool dropping db node
Date: Wed, 13 Feb 2008 17:31:41 -0200

> I have one pgpool-II instance, running with 2 backend postgresql
> servers, using replicate_mode with load balancing.
> First, and most important problem, is the fact that after some time (I
> guess this happens when load is high), it drops the second backend. I
> couldn't find a way to make pgpool log things. The documentation says
> all it logs is its pid in /var/run.

"logdir" parameter means that pgpool puts on a pid file in the
specified directory. If you want to collect pgpool's log, you need to
start up with the following command.

  % pgpool -n > /var/log/pgpool.log 2>&1 &

or

  % pgpool -n 2>&1 | rotatelogs /var/log/pgpool 86400 &
  

> When this happens, I have to go all the way to sync the backends and
> restart the pooler. I do this by stopping the pooler (normally it
> won't stop, and I must use killall -9 on it). After I kill the pooler,
> I stop the backends, rsync them, and then start the pooler again after
> starting the backends. Sometimes, pgpool won't start, sometimes it
> starts and then dies, and sometimes it starts fine. I noticed the
> problem of not starting got worse after I increased the number of
> pre-forked childs.
> Is there any way in which I can help debug this?

Hmm... Could you show us debug log, stack trace and result of ps command?

  % gdb -p <pgpool pid>
  (gdb) bt

  % ps auwwx | grep pgpool

Regards,
--
Yoshiyuki Asaba
y-asaba at sraoss.co.jp