Sometime ago i was looking for something like this and because at this time
XC was a little baby i tried installing bucardo but i gave up when stucked
fighting with perl modules. So, after testing some other solutions i decided
to make my own, just "touching" the trigger part of the pyreplica project
and building a daemon in a binary compiled program (no script) for a better
performance. It's working well by now, with some tables and few nodes (12
nodes deployed across long distances), but i was lucky then because the
database and the systems were still in design and development phases.
There are many things you have to keep in mind, as all people is warning
you. I tell you, it can be done, but you MUST redesign your existing
database preparing for multimaster paradigm where the key is avoid node
level concurrency, and your client database apps to THINK what to do when
nodes are down, but if you are not ready to prepare your database (design)
or you have not access to the source code of your client database apps then
you have problems.
I suggest you read this
http://www.dbspecialists.com/files/presentations/mm_replication.html and
this
http://scale-out-blog.blogspot.com/2012/04/if-you-must-deploy-multi-master.html.
Finally, i can mention another posible solution that i never tried but maybe
can help you, SymmetricDS.
Good luck.
--
View this message in context: http://postgresql.1045698.n5.nabble.com/Multi-Master-Replication-tp5783855p5784468.html
Sent from the PostgreSQL - general mailing list archive at Nabble.com.
