[pgpool-general: 3844] Re: health check on standby node

[Yugo Nagata]

On Wed, 1 Jul 2015 06:21:27 +0000
Wes Mitchell <wes.mitchell at ericsson.com> wrote:

> I am puzzled by the following behavior:
> 
> 
>   1.  Two nodes, each running pgpool/PG, master on 0, hot standby on 1, configured for streaming replication.
>   2.  I bring down the database on the standby, node 1
>   3.  Both pgpools mark the standby node as down
>   4.  I bring the standby database back up
>   5.  Nothing happens

> 
> If I manually restart the standby, should I always use pcp attach node to get the pgpool processes to recognize the standby?

Yes, you should use pcp_attach_node or pcp_recovery_node for geting
a backend back to under pgpool-II, or restart pgpool-IIs with -D option.

> Or, is there a config parameter I can set to have pgpool try the dead standby periodically, to detect when it comes back?

No. Once a backend is dettached, this never get back to pgpool-II
automaticaly. This prevent any problematic backend getting back
to cluster.
> 
> Thanks
> Wes


-- 
Yugo Nagata <nagata at sraoss.co.jp>