[pgpool-general: 7337] Re: Force failover of pgpool watchdog leader

[Emond Papegaaij]

Hi Tatsuo,

Shall I file a feature request at the bugtracker for this?

Best regards,
Emond

On Thu, Nov 12, 2020 at 8:00 AM Tatsuo Ishii <ishii at sraoss.co.jp> wrote:
>
> > Hi all,
> >
> > Is it possible to force the watchdog leader to switch to a different
> > node in the cluster without stopping the current pgpool leader process
> > entirely?
> >
> > The reason I want to do this is to automate switchover when a host in
> > my cluster is shut down or rebooted. Every host runs a database and a
> > pgpool instance. So if the machine that is about to shut down is both
> > pg primary and pgpool watchdog leader, both need to switch over to a
> > different node. I can detach the pg primary to force a switchover, but
> > if I then shutdown the pgpool leader (running on the same host), the
> > database switchover is not completed correctly. When a different host
> > is watchdog leader, the database switchover process continues when the
> > shutdown host disappears and my cluster remains consistent.
> >
> > My current solution is to stop pgpool if it currently is the watchdog
> > leader, start it again, wait for some other node to pick up the leader
> > status, trigger the database switchover and finally shutdown pgpool
> > and postgres. This process is lengthy, complex and increases the
> > amount of time the node is in a broken state.
>
> Unfortunately there's no such a feature in watchdog. Maybe it's better
> to add this request to the TODO list (for 4.3 development).
>
> Best regards,
> --
> Tatsuo Ishii
> SRA OSS, Inc. Japan
> English: http://www.sraoss.co.jp/index_en.php
> Japanese:http://www.sraoss.co.jp