Setting up PGCluster

From Alpine Linux
Revision as of 18:52, 5 June 2009 by Cbanta (talk | contribs)
Jump to: navigation, search

PGCluster is a fork of PostgreSQL that allows multi-master replication. This document explains how to set it up.


  • All hosts in conf files are referenced by name - these must resolve properly to thier IPs.
  • PGCluster xml conf files are case sensitive
  • PGCluster 1.9.0 is based off PostgreSQL 8.3


First install the packages

apk add pgcluster pgcluster-client

For recovery, you'll need ssh and rsync

apk add openssh rsync


Edit /etc/hosts file if necessary so all names can be resolved.

Check the data path in /etc/conf.d/pgcluster. Also check here to see which processes the init script will start on this host.

Create the database files (same as for postgresql)

/etc/init.d/pgcluster setup

Edit the config files in the data path found above (usually /var/lib/posgresql/pgcluster)

postgres (postmaster)

postgres has 3 configuration files of interest here: postgresql.conf, pg_hba.conf, cluster.conf.

To start it manually:

su postgres
cd /var/lib/postgresql/pgcluster
pg_ctl -D . start


postgres -D .


This is the standard postgresql config file. Just make sure of this:


You might also want to changed the logging settings.


This is also a standard postgresql config file. Make sure all the hosts involved trust each other. Here's an example line you can use to get a test cluster running (it's a bit permissive for production use):

host    all         all        trust


This file is new with PGCLuster - it configures the patched database instance.

Add a block like this for each replication server:

	<Host_Name>     pgrep1  </Host_Name>
	<Port>          8001    </Port>
	<Recovery_Port> 8101    </Recovery_Port>

Also add the following:

<Host_Name>		pgdb1				</Host_Name>
<Recovery_Port>	7001				</Recovery_Port>
<Rsync_Path>		/usr/bin/rsync			</Rsync_Path>
<Rsync_Option>	ssh -1				</Rsync_Option>
<Rsync_Compress>	yes				</Rsync_Compress>
<Rsync_Timeout>	10min				</Rsync_Timeout>
<Rsync_Bwlimit>	0KB				</Rsync_Bwlimit>
<Pg_Dump_Path>	/var/lib/pgcluster/pg_dump	</Pg_Dump_Path>
<Ping_Path>		/bin/ping			</Ping_Path>
<When_Stand_Alone>	read_only			</When_Stand_Alone>
<Replication_Timeout>	1min				</Replication_Timeout>
<LifeCheck_Timeout>	3s				</LifeCheck_Timeout>
<LifeCheck_Interval>	11s				</LifeCheck_Interval>

And this bit is optional:

	<DB_Name>		test_db		</DB_Name>
#	<Table_Name>		log_table	</Table_Name>


This process replicates to the other servers.

To start it manually:

su postgres
cd /var/lib/postgresql/pgcluster
pgreplicate -D . 

If you want debug output change the last line to:

pgreplicate -D . -n -v


Add a block like this for each database (also called cluster) instance:


Add this block for your load balancer instance:


And add this to describe the replicator instance you're configuring:


Optional Logging info:



This process provides load balancing.

To start it manually:

su postgres
cd /var/lib/postgresql/pgcluster
pglb -D . 

If you want debug output change the last line to:

pglb -D . -n -v


Similar to pgreplicate, add the following block for each database instance:


Add the following for the load balancer:


Optional Logging info: