[pgpool-general: 3107] Re: Possible split-brain scenario

Tatsuo Ishii ishii at postgresql.org
Wed Aug 20 08:26:03 JST 2014


> I'm trying to figure out if pgpool can avoid a split-brain scenario. Say my
> primary server loses internet connectivity including with both my standbys, but
> those standbys are still happily communicating with each other and with the
> network. In that event, I would expect the first standby to promote itself to
> become the new master and the second standby would start following it. I've
> actually seen this work here. But, what happens to the original primary in the
> event that network connectivity is restored? Unless it is demoted to become a
> standby to the new primary, which in fact may not succeed, then I have a
> split-brain situation. Would pgpool degenerate the old primary on reconnection?
> I'm trying to figure out which pgpool would now have control of the overall
> system: the one on the old primary or the one on the new primary?

In this case old primary is treated as "down" state by pgpool and
pgpool does not route any request to the down node. So as long as
clients connect to pgpool to use PostgreSQL, there should be no
problem (no split brain). Of course if a client connects to the old
primary directly (not via pgpool), that will be a problem but it's
user's fault.

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


More information about the pgpool-general mailing list