[pgpool-general: 6763] Re: Watchdog does not get consensus for failover

[Tatsuo Ishii]

>> Thanks. I confirmed that pgpool at proxy2 and proxy3 did not detect
>> the failure of postgres1. So watchdog at proxy1 did not get
>> consensus. I suspect the way you stopped postgres was something special
>> (not ordinary shutdown by using pg_ctl). Can you elaborate how you
>> stopped postgres1?
>>
> 
> I used "service postgresql stop" on postgres1. And then "start" to start it
> again.
> 
> So I have one question: why you ask me this? Does this mean that if
> PostgreSQL would be killed in an "abnormal way" (like kernel hang, memory
> or something else not made by user) the failover does not work?

No, in these cases Pgpool-II's health check should detect the
failure. I thought about the case a firewall was tweaked to block
access from pgoool/proxy1 to PostgreSQL.

Anyway, the debug log and "show pool_nodes" results look pretty
strange since pgpool/proxy2 and pgpool/proxy3 thinks that PostgreSQL
is healthy, while pgpool/proxy1 does not. For now I have no idea what
could cause this situation.

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