How does ejabberd clustering work conceptually?

Question

I am a complete newbie to XMPP and ejabberd. I would be very thankful if some expert could tell me, how the ejabberd "clustering" works. I searched a lot, but could not find any specs about this.

Does it simply replicate all it's data via http (so it doesn't matter where a client connects)? Or will it be done via DNS RR?

Answer 1

It seems that clustering is done on an DNS Level (or frontend LB) and the basis being a clustered database / persistence layer. They provide several implementations for this.