Proper implementation for hot fail over load balancing zimbra

[Everyday]

We're revisiting zimbra once again but I wanted to see how others are implementing hot fail over and load balancing. We are currently running kerio because it supports true hot fail over. Without getting into a lot of useless talk I will simply ask...

What is the preferred way to setup a load balanced zimbra cluster with hot fail over? When I say hot fail over I mean if one of the servers in the cluster goes down there is no interruption in service to any user. By load balancing I mean no one server will be overloaded with users and it does not require manually balancing new account creation across the cluster.

Any and all input is appreciated.

[Klug]

Zimbra is not designed for load balancing. At all.
You can do HA (two nodes and share volumes) but not load balancing.

[Everyday]

When doing HA we would still need to manually switch records within zimbra whenever a server might go down?

[nrc]

For load balancing check the multi-server installation guide.

http://www.zimbra.com/docs/os/latest...1.html#1074736

You can't do round robin or load based load balancing but you can break up your users into different mail stores and use that to break up your user load. Of course that means that you need failover for each mail store.

We plan to use VMware to keep hardware failover out of the equation - at least as far as our Zimbra installation is concerned.

[Klug]

Originally Posted by Everyday

When doing HA we would still need to manually switch records within zimbra whenever a server might go down?

No, because Zimbra's services bind to a virtual IP that "sticks" to the active node.