Recently I ordered 2 PowerEdge r710 Servers with a PowerVault MD3000i shared disk array. Each servers had 2 x 2.53GHz Quad Core Intel Xeon E5540 Processors and 32 GB RAM.
The idea is to virtualise the entire Harrogate Datacenter that we have, to:
- Save on Power
- Consolidate out of date hardware
- reduce costs
- provide a High Availability system to our End Users and clients who will log in via our extranet.
I have only until now installed proof of concept and test labs with Microsoft Hyper V and Clustering using 3 Optiplex Desktops (2 for the cluster nodes and one with windows 2008 storage server for the shared storage).
I could not believe how simple it was to actually setup and configure the cluster and then deploy Hyper V on to it.
The live environment will be completely resilient, 2 connections to the iSCS SAN Dual Power etc so there will no one single point of failure. As I am only using 2 nodes I can connect the SAN directly to the servers as there are 4 iSCSI ports on the MD3000i, had I had more nodes then I would need to use 2 switches dedicated for the iSCSI SAN.
Once the servers were installed I then setup the clustered environment. The cluster wasn’t only for HyperV but for SQL as well.
I had some issues installing the SQL cluster initially, turned out I needed to slip stream SP1 in to SQL 2008 in order to install it on a cluster. Once SQL and HyperV were installed on the server I was ready to start deploying guests on to the server.
The difference with HyperV on Server 2008 R2 from Server 2008 is that you can benefit from “Live State Migration” you can move a guest from one node to another with 0 down time on the guest, much like the V-Motion that you get with VMWare. In order to make use of the Live State Migration I needed to assign a virtual disk to each guest if you have more than one guest on a virtual disk the live migration will fail.
Again until now I would always create the virtual disk on the cluster and then format it and assign a drive letter to it, I then thought what if I went over 23 guests? how would it then work once I have reached the end of the alphabet?
The answer was simple, don’t assign a drive letter to the Virtual Disk, I wouldn’t have actually thought about that if I had not been “playing around” with Ubuntu recently, as linux does not assign a C:\, D:\ etc to disks. OK so if I do not assign a letter it assigns a rather long ID to the disk but you can suss it out when it comes to adding the guest on the host as to what disk is for what guest.
Also up until now I had not used any management tools for HyperV except the standard HyperV Manager that is installed initially. So I downloaded Microsoft Virtual Machine Manager Console 2008 R2 (MSVMMC) and started looking at this. I installed this as a guest on the Cluster. I have to say that I would now be lost without MSVMMC as it is a vital tool to managing HyperV servers regardless of being in a cluster or not. There are a lot of tasks that can be done with it that you could not do with the standard HyperV manager one for example is to create a virtual from a physical (P2V) Very useful indeed!
I am still learning about the management of the environment but at present I have a very stable and highly available environment for the external facing servers. I have recently started at looking at virtualising our Edinburgh office server infrastructure as we can possible consolidate 20 servers to a 3 node clustered HyperV environment.
I know that many of my peers in other firms would have considered and some already are deploying VMWare, however As we are already licensed for HyperV and Clustering with our Server 2008 Ent R2 licenses I cannot see why we need to look at another alternative as this is the most cost effective way of moving forward. The next step forward for me is to look at Windows Server 2008 R2 Datacenter I believe that we can step up our current Windows Server Std and Ent licenses to Datacentre which would then mean we would be licensed to run and unlimited number of Windows guests on our HyperV cluster.