[pgpool-general: 1952] Re: pgpool 3.2.4: Watchdog doesn't start if not all "other" nodes are alive

Yugo Nagata nagata at sraoss.co.jp
Mon Jul 29 11:45:46 JST 2013


On Fri, 26 Jul 2013 10:49:18 +0200
Sam Wouters <sam at ericom.be> wrote:

> Hi,
> It took me a while to found out what was wrong and I didn't find any 
> related issues on the web, so I thought I should share;
> I'm setting up a three node pgpool cluster (pgpool running on every 
> database server) and noticed that the watchdog process was stuck after 
> wd_init. As a result, failover of the pgpool didn't work.
> The issue was I was setting up the cluster and didn't setup the third 
> one yet, but already included it in the pgpool.conf. Apparently the 
> lifecheck cyclus won't start if at startup time one of the nodes is not 
> available.
> Don't know if this is intended behaviour or changed in newer versions, 

This is intended and not changed in the new version.

This is written in "Starting/stopping watchdog" section of the manual.

  "The life checking starts after waiting for all of the pgpools to start. "

(However, his is so confusing. I'll rewrite it.)

> but I would think that it would be better to assume nodes are down at 
> startup, initiate some learning cycles and set there correct status 
> after that.

Thanks for your proposal. We'll discuss about this.

> regards,
> Sam

Yugo Nagata <nagata at sraoss.co.jp>

More information about the pgpool-general mailing list