[pgpool-general: 2298] Re: node status after restart

[Tatsuo Ishii]

> Hello list,
> 
> I think I have something configured incorrectly.  I have postgres 9.2.5
> with pgpool 3.3
> 
> I have postgres setup with streaming replication, and pgpool in
> master/slave mode.  I have a db1 and a db2 (2 nodes).  When i have db1 as
> the primary and db2 as the secondary, and I shutdown db1, it fails over as
> expected and db2 becomes primary.  I have a problem when I restart db1.  I
> expect pgpool to see that it is back online and for it to come back as a
> standby.  It doesn't do this and stays in the down status according to
> pgpool

pgpool does not failback the previously downed node (in your case db1)
automatically.  This is intended behavior because db1 is not
inconsistent against db2 anymore, due the difference of timeline.

PostgreSQL 9.4 maybe able to overcome the limitation, but not now.

> 2013-11-15 20:08:10 DEBUG: pid 12634: health_check: 0 th DB node status: 3
> 2013-11-15 20:08:10 DEBUG: pid 12634: health_check: 1 th DB node status: 2
> 
> Are there any settings I should be looking at?  I don't know why pgpool
> doesn't see the database is back online and ready to be recovered.

You need to execute online recovery to make db1 online state. For this
purpose you need to setup online recovery script. See manual for more
details.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp