On Thu, Apr 14, 2011 at 08:06:28PM +0800, Yogi Rizkiadi wrote:

> i want to implement project with database replication structure 3 layers : A->B->C, which is A is the master node, B are 60 nodes of datacenter, and C are 1 - 7 nodes of machines on every datacenter. So if we count it has more than 100 nodes.
> 
> Every machines on layer C only have 1 GB of RAM and network on 3 layers above are perhaps unconsistent.
> 
> Is it possible to use Slony-I 2.0.4 and where can i get a good example for doing that ?

I think Slony's overhead is going to kill you with that many nodes.
You might be able to segment up the network such that it wasn't too
bad (especially if many of them can't forward), but I'd be sceptical.

You might try chaining things, though.  I have wondered about a
two-layer Slony replication system with a bunch of WAL-streaming
replicas hanging off the second layer.  You have of course various
single points of failure, but it's possible to recover in most cases.

A


-- 
Andrew Sullivan
ajs at crankycanuck.ca