[pgpool-general: 2992] Re: Confusion between backend and Watchdog nodes?

Yugo Nagata nagata at sraoss.co.jp
Mon Jun 30 20:33:36 JST 2014


pgpool's active/standby switching occurs because active pgpool 
fails to issue lifecheck query "SELECT 1" to itself before
failover of backend finishes.

To avoid this, health_check_period should be smaller, or
wd_interval and wd_life_point should be larger.

In fact, I recommend to use 'heartbeat' mode rather than 'query'.
The query mode can be problem because pgpool failure and backend
failure are not well seperated.

On Wed, 4 Jun 2014 10:44:07 -0500
Jeremiah Penery <jpenery at redskytech.com> wrote:

> Hi,
> I'm Using pgpool-II 3.3.3 and pgpoolAdmin 3.3.1, and I have a setup with
> 2 pgpool nodes and 3 backend database nodes using streaming replication.
> I start everything up and pgpool node 0 is the active Watchdog node, and
> backend node 0 is the primary database node.
> When I try to test failover by rebooting database node 0, the database
> failover works fine, but pgpool also considers Watchdog node 0 to have
> gone down as well.  The only reason I can come up with is that it's
> somehow linking the backend node numbers to the watchdog node numbers,
> which seems very strange.  I've been able to reproduce this
> consistently.
> I have attached the pgpool logs from both servers, as well as the
> configuration from the pgpool 'primary' node.
> Is there something I'm doing wrong in the way this is configured?
> Thanks,
> Jeremiah

Yugo Nagata <nagata at sraoss.co.jp>

More information about the pgpool-general mailing list