Windows 2008 Multi-Site Failover Cluster with 2 Virtual ESX4 Nodes

have you ever got an answer of this, if yes, can you please share your experience on setting up Multi-Site Failover Cluster by using MV/CE and virtual w2k8 servers…

Just a thought - when you have mirrorview running, you can not access the mirror on the DR site unless you promote the mirror. Mirrors are designed for disaster recovery, not clustering - not sure if this was part of the question or not.

glen

Hi,

Yes I got it working perfectly well in the end. The problem I had was with Windows 2008 R2 in that EMC MV/CE does not support R2 yet, so I had to drop back to 64-bit R1 release of Windows 2008.

It takes approximately 3 minutes for MV/CE to failover to the node at the other Data Centre. This includes the time to promote the Mirrorview secondary copy to the primary image, rescan disks on the vSphere hosts and reverse the MirrorView synchronisation. It has worked 100% every time and it always takes less than 3 minutes 15 seconds to complete the task.

So in summary of my functional setup – I have 2 virtual machines. One VM is located in each Data Centre (connected by dark fibre). I have 2 x EMC CX4 arrays, one at each data centre with MV/S and MV/A licensing. I have 4 x LUNs presented to each VM as RDM’s. The primary data centre is the primary MirrorView image of each LUN, with MV/S replication to the DR data centre. Those secondary images are presented as RDM’s to the VM at that DR site.

Solutions Enabler is installed on both VM nodes, as is MirrorView/CE. The cluster is configured using Windows’ Failover Cluster service. NOTE: You HAVE to add the disks using the “cluster” command line as it will not find the disk using the GUI. The cluster itself is setup as a Majority Node Set with File Share Witness, so I have a third server (one of my fibre-connected sites – NOT one of the Data Centres themselves) that has the file share witness share on it. Once the cluster is setup, I used MV/CE management tool to configure the CE cluster. Basically it sets up a consistency group with the LUNs of the cluster, and it sets up communications with both arrays so it can do the fracturing and promoting etc. It also adds a resource to the cluster group, which is basically the “trigger” for MV/CE to do the array tasks during failover.

It has been running for about 2 or 3 months now and we’ve failed it over and back several times (both intentionally and due to failures) in that time.

Regards,

Bobby Stojceski VCP3/4

Infrastructure Manager

Transpacific Industries Group

Bobby,

Thanks very much for the very detailed solution that you proivded - this should be a great help to others facing the same issues that you did.

Please mark  your question as answered and give youself all the points for this solution. 

glen

great info, thanks a bunch…

are the failover process all handled by the MC/CE failover cluster manager automatically?

do you have to manually scan the esx host in DR site to recognize the disk?

Everything is automatically done by MV/CE. Nothing manual is required. It does the rescans on its own at both nodes.

It has enough smarts in it not to failover if mirrorview hasn't been replicating (broken link). But you can override that if you wish.

How do I give myself points and close the discussion? Couldn't see that anywhere

Just joking - you can only give others points - but in your case you deserve the points but no way to do that 

glen

fyi

MirrorView/CE version 4.1 was released on Dec.,16th 2010, I think it supports 2008 r2 now.