Hi Jen,
On Tue, 15 Oct 2013 21:09:50 +0000, Jeni Fifrick <jfifrick@incomm.com>
wrote:
> We're having Master & Slave Postgres 9.1.3 on CentOS - using
> Streaming Replication, which serve as a read-only database, and as
> 'backup' for the production database.
Sure.
> I'm trying to setup a process that can be used to have the continuous
> backup of the production, to make sure if some problems happen on the
> production server, we won't loss the data.
It is important that you understand though that such an architecture
(with just replication) won't completely protect you from disasters
happening on the master. I am referring to data loss caused by
undeliberate DELETE or DROP operations, for example, or SQL injections.
Especially if you have a streaming replicated standby, data that
disappears on the master will disappear from the standby immediately
after.
Physical backup and continuous archiving will add this kind of
protection to your system. Using 9.1 unfortunately you can't benefit
from cascading replication using the streaming protocol (introduced in
9.2), however you can look into the pg_basebackup utility for this
purpose.
Otherwise, there are tools out there that help you achieving these
goals, such as OmniPITR (https://github.com/omniti-labs/omnipitr),
pg-rman (http://code.google.com/p/pg-rman/), WAL-E
(https://github.com/heroku/WAL-E, mainly for Amazon AWS users) and
barman (http://www.pgbarman.org/) - to name a few.
Being one of the developers, I can definitely recommend Barman. It does
not yet support streaming replication, but usually with WAL file
shipping you can still achieve very good results in terms of Recovery
Point Objective (which measures data loss). It is very easy to setup,
and already has packages for CentOS.
I hope this helps.
Cheers,
Gabriele
--
Gabriele Bartolini - 2ndQuadrant Italia
PostgreSQL Training, Services and Support
Gabriele.Bartolini@2ndQuadrant.it - www.2ndQuadrant.it