When demonstrating Windows Server 2012 (WS2012) Hyper-V Replica (HVR), one of questions I was asked most often was “can I replicate the virtual machines to a third site?” In the case of WS2012 Hyper-V, the answer was no – you can only do A-B replication from Site A to Site B. However, Windows Server 2012 R2 (WS2012 R2) gives us a new feature called Hyper-V Replica Extended Replication, which I will explain in this post.
In WS2012 you can replicate a running VM from Site A to Site B, otherwise known as A-B replication. There is no option to replicate the VM to a Site C. This limited replication patterns, as you can see in the following examples:
WS2012 R2 Hyper-V brings extended replication. This allows us to do A-B-C replication. This should be read as A-to-B replication followed by B-to-C replication. What does this mean you can do?
Extended Replication in WS2012 R2 Hyper-V does not give you A-B and A-C replication where a VM is directly replicated from Site A to Site B and Site C simultaneously.
What solutions could you implement with Extended Replication? The objective of this feature is to give Site B extended protection after failing over from Site A to Site B. Here are a few ideas:
Site A (Primary Site) | Site B (Secondary Site) | Site C (Third Site) |
Campus computer room 1 | Campus Computer room 2 | Remote owned DR site |
Campus computer room 1 | Campus Computer room 2 | Public cloud DR site |
Branch Office | Local DR site | HQ DR site |
Production site | DR site | Public cloud DR site |
Extending Hyper-V Replication from Site B to Site C.
In other words, if you can put WS2012 R2 Hyper-V hosts/clusters into a third site and enable Hyper-V Replica from the second site, then you can extend replication from Site B to Site C.
The first step is to enable Hyper-V Replica from Site A to Site B. That’s just a basic configuration of Hyper-V Replica, maybe adding in usage of the Hyper-V Replica Broker. You will extend replication in the DR site, Site B:
You have to consider what impact Extended Replication will have on the Recovery Point Objective (RPO) between Site A and Site C. If Site A replicates to Site B every 30 seconds, and Site B replicates to Site C every 15 minutes, then there is a 15 minutes and 30 seconds gap between Site A and Site C. In other words, if you lose both sites A and B in an unexpected disaster, then you lose fifteen and a half minutes of data if you restore the newest recovery point in Site C via an unplanned failover.