Thread: pgpool2 master fail problem

pgpool2 master fail problem

From
Videanu Adrian
Date:
Hi all,
i try to build a postgresql cluster using postgresql 9.1.3 and pgpool2 3.1.2.
I want to setup this cluster as a master/slave with streaming copy. Everything is ok execept the case when the masters
fails,and i try to recover it, because the old master recovers as master instead of slave. I have installed
pgpool-walrecrunningon both servers but still does not seems to work. 
any ideea ?

Re: pgpool2 master fail problem

From
Tatsuo Ishii
Date:
> Hi all,
> i try to build a postgresql cluster using postgresql 9.1.3 and pgpool2 3.1.2.
> I want to setup this cluster as a master/slave with streaming copy. Everything is ok execept the case when the
mastersfails, and i try to recover it, because the old master recovers as master instead of slave. I have installed
pgpool-walrecrunningon both servers but still does not seems to work. 
> any ideea ?

I'm not sure I fully understand your complain but I guess you expected
the failed master became new standby by just restarting it? That is
not allowed with streaming replication system anyway. You need to
recreate the failed master as a new standby from base backup. Pgpool
provides a convenient way called "online recovery" to do that.  See
pgpool manual or tutorials at pgpool web site (http://www.pgpool.net),
especially this one:
http://www.pgpool.net/pgpool-web/contrib_docs/simple_sr_setting2_3.1/index.html
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp

Re: pgpool2 master fail problem

From
John R Pierce
Date:
On 04/17/12 7:10 AM, Videanu Adrian wrote:
> i try to build a postgresql cluster using postgresql 9.1.3 and pgpool2 3.1.2.
> I want to setup this cluster as a master/slave with streaming copy. Everything is ok execept the case when the
mastersfails, and i try to recover it, because the old master recovers as master instead of slave. I have installed
pgpool-walrecrunningon both servers but still does not seems to work. 

in addition to the streaming replication built into postgres, you need
some form of cluster management software running over the top of this to
manage the transition states, ensure that you don't have two masters,
etc.    as another said, the previously failed master has to be brought
online as a new slave

--
john r pierce                            N 37, W 122
santa cruz ca                         mid-left coast