PostgreSQL replication strategies

Question

Right now we are using PostgreSQL 8.3 (on Linux) as a database backend to our Ruby on Rails web application.

Considering that on PostgreSQL database we actively use row level blocking and PL/PGSQL, what can we employ to secure our data -- I mean tools, packages, scripts, strategies -- to successfully replicate the database and build multi-master combination?

I will appreciate master-slave suggestions as well.

For example, if I put several application servers running Apache/Ruby to achieve higher performance and at the end deploy several database servers, is there any way to build multi-master replication in PostgreSQL?

Right now we use PostgreSQL WAL mechanism to backup data to file system.

Thanks a lot.

Answer 1

There are a few tools for master-slave (and master-multislave) scenarios, usually trigger-based. Slony-I has already been mentioned (is stable and solid, but a bit difficult to operate). People having problems with Slony-I wrote Londiste (by Skype team) and PyReplica. Bah, and I just spotted Mammoth has been open-sourced

For multimaster there is Bucardo (note: it is not that polished) or commercial offerings - for example by Continuent or CyberTec.

Answer 2

If you haven't already, I'd suggest a look at the High Availability, Load Balancing, and Replication chapter of the PostgreSQL manual. It gives a clear overview of the available techiques and their features.