[pgpool-hackers: 3409] Re: watchdog behavior regarding VIP

[Tatsuo Ishii]

> Hi Ishii-San,
> 
> On Tue, Sep 3, 2019 at 5:43 AM Tatsuo Ishii <ishii at sraoss.co.jp> wrote:
> 
>> Hi Usama,
>>
>> Can you comment on this:
>> https://www.pgpool.net/mantisbt/view.php?id=541
>>
>> It seems VIP is not set when there's no quorum. Since there's nothing
>> in the docs regarding VIP vs. quorum and I am wondering if this is
>> a feature or a bug.
>>
> 
> This is a feature and as per the design. The reason for doing this is to
> avoid split-brain syndrome.
> In the absence of a quorum, the watchdog cluster does select a master node
> but waits for the quorum to perform the escalation(bringing up of VIP).
> 
> Initially, in the absence of the quorum, the cluster used to abstain from
> selecting the master node but it was causing some synchronization issues so
> the behaviour was modified and now the cluster goes on the elect the master
> node even in the absence of the quorum but does not bring up the
> VIP until it finds the quorum.
> 
> Below mentioned commit changed the said behaviour back in 2015
> 
> https://git.postgresql.org/gitweb/?p=pgpool2.git;a=commitdiff;h=6238ba9d2aaa82db415b11aca90d54981a854e09

Thanks for the explanation. One more question.

Suppose 3 node watchdog configuration. Initially all 3 nodes are up
and VIP is up. Now 2 of 3 nodes go down, and quorum is lost, so the
VIP will be brought down?

Best regards,
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp