Base information for both primary and standby


Primary Server

Create user dedicate for replication

--you can specify any user name and password
--the 'replication' privilege is important
create user r replication password '1';

Grant connection accessibility

After creation of user, you need to modify the $PGCONF/pg_hba.conf file to explicitly grant connection accessibility for the user r

#You can choose any of the authentication method you like
echo "host    replication    r    $standby_ip/32    md5" >> $PGCONF/pg_hba.conf

WAL setting

According to Official Document, we need to set wal_level to hot_standby so that PostgreSQL will generate WAL that contains enough information for standby to reconstruct the status of running transactions from the WAL.
We also need to set max_wal_senders to a positive number to accept connection from replicators. This is the number of sender process to send streaming data, so it is better to put a bigger number like 5 or 10.

echo "wal_level = hot_standby" >> $PGCONF/postgresql.conf  
echo "max_wal_senders = 3" >> $PGCONF/postgresql.conf  

Restart primary server

To enable streaming function, we need to restart Postgresql server, this will also reload the pg_hba.conf so that remote replicator could connect to this server.

sudo service postgresql restart

Now your primary server is ready to accept streaming replication connection.

Extra notice

For beginner who does not know about connecting to remote server, please refer to this article.

Standby Server

Restore base data files

To achieve this, it is better to use pg_basebackup as it is more convenient than scp them manually.

pg_basebackup -U r -h $master_ip --format=plain -vxP  -D  $PGDATA

After a while, you will find the primary server data folder is replicated to standby server.

Recovery configuration

The most important configuration comes, you need to create a recovery.conf file in you $PGDATA folder.

#Notice the single quote around (on)
#This will instruct this server to do standby work
echo "standby_mode = 'on'">>$PGDATA/recovery.conf
#The libpg connection string specified here allows standby connecting to primary
#Hence those information should be consistent with primary
echo "primary_conninfo = 'host=$master_ip port=5432 user=r password=1'">>$PGDATA/recovery.conf

Up to now, the standby server could start replication if you restart it. But this standby server is not connectable, which means no connection allowed to this standby server, thus if you try to connect to it, you will get:

psql: FATAL:  the database system is starting up
FATAL:  the database system is starting up

But be easy, even it is not connectable from you, it could replicate content from primary and also dispatch them to slavers depending on your configuration.

Hot standby server

So to allow connections to standby server rather than leave it unconnectable, you need to modify or add one configuration in standby server, refer to Official Document:

echo "hot_standby = on" >> $PGCONF/postgresql.conf

You can find this entry under standby section. This attribute means to allow connection and run read-only queries while the server is in archive recovery or standby mode. In our case, in standby mode but accepts read only connection.
Such configuration could also be used as failover setting, which I will introduce later on.


Now try to create some tables or do some DML on Primary server and see if they will be synchronized to Standby server.


Those configuration above are the minimum for launch a streaming replication between primary and standby.
For more high availability information, please refer to failover, WAL archive, HA.
I will keep posting tutorial.


20 July 2015