How Microsoft Hyper-V and the Cisco UCS changed our lives

Cisco UCS Hardware

At the end of last year we had our Cisco UCS ordered and in your datacenter. In January we started the testing and made the Clusters ready for the production environment. In February we started the migration of our existing environment, mostly P2V and also some V2V migrations.

Here some interessting facts about our Cisco UCS and Hyper-V project.

  • We use 12 Cisco UCS Blades this is like 10 HE of rackspace
  • We migrated 45 Windows Servers and 47 Unix Servers in just one week
  • We replace 2 racks of server with a half rack of two Cisco UCS Bladecenters
  • We think we can replace 2-3 racks more with our two Bladecenters.
  • At the end of this year we think we could replace 4-5 racks with 1/2 rack
  • We still have a lot of physical and virtual server which will be needed to be migrated to the Bladecenter.
  • We will get even more out of our Blade Servers by activting Hyper-V Dynamic Memory as a new feature of Hyper-V R2 ServicePack 1

This migration had a lot of positive influence on other things in the datacenter.

Datacenter Power

  • we need now 4% less energie overall
  • we need now 6% less cooling overall
  • we need less space (1 and 1/2 racks at the moment)
  • now our system administrator travel 50% less to the datacenter, because of hardware defects or other administrative tasks.
  • We can deploy new servers in minutes instead of hours

I think all of this numbers (except the time we need to deploy new servers )will increase after the next migrations.

Now I started to write a series of blog posts about installing Microsoft Hyper-V R2 on the Cisco UCS system:

Microsoft Hyper-V and the Cisco UCS Bladecenter are a powerful team. The UCS Virtual Hardware takes alot of complexety from the hypervisor in your case Hyper-V. You don’t need NIC teaming and stuff like that. Thats is making it very easy to deploy Hyper-V Clusters. And with the Microsoft System Center Virtual Machine Manager you can save a lot of time in managing your clusters, hosts, virtual machine and also in P2V and V2V migrations. Since Microsoft SCVMM supports Windows Powershell you can also do a lot of scripting automation. And with the release of the new Version of SCVMM (System Center Virtual Machine Manager 2012) this will even get better for deploying new virtual machines, services and even public and private clouds.

Hyper-V R2 SP1

We started with Microsoft Hyper-V R2 Servers before the release of Service Pack 1. We think we can even get a lot more out of your systems with the new Dynamic Memory feature for Hyper-V which comes in Service Pack 1.

At the end we think choosing the Cisco UCS, Microsoft Hyper-V and Microsoft System Center Virtual Machine Manager for our datacenter was the best choice we have made, in terms of costs and technology.

Cisco UCS Hyper-V Cluster – Important Updates for the Hyper-V Cluster – Part 9

Since we have installed our Microsoft Hyper-V Cluster on the Cisco UCS, Microsoft released some patches for Hyper-V, Windows and Clustering.

There are two really important Updates which I would recommend for Hyper-V Clusters.

  • The first is Service Pack 1 for Windows Server 2008 R2 and Hyper-V Server 2008 R2. Service Pack 1 brings a lot of Hotfixes for Hyper-V, Failover Cluster Feature and other Microsoft Server features. And it brings also a two new features called Dynamic Memory and RemoteFX.
  • The second one is a hotfix for Servers with Intel Westmere or Sandy Bridge and has a large amount of physical memory. Most of the Cisco UCS Blades will meet this configuration. You can get more information on this Hyper-V hotfix here.

This two updates will bring you a much better experience with your Hyper-V Cluster. It will improve performance, stability and it will add new features.

Hyper-V Hotfix for Intel Sandy Bridge and Westmere

If you are using Microsoft Hyper-V with new Intel Sandy Bridge or Intel Westmere processors and our server has over 48GB of physical RAM you may get a performance decrease. Microsoft released a hotfix (KB2517329) for this problem.

This could also be a problem with a lot of Hyper-V installations on Cisco UCS, because the most of them will use more than 48GB of memory.

The performance decrease could show its self in the the following issues:

  • The CPU usage is high and the server responds slowly when you copy large files on the computer. For example, you copy a 10-GB file
  • The disk I/O performance of the virtual machines (VMs) is slow
  • Windows takes a long time to start

The hotfix is not include in Service Pack 1 for Hpyer-V R2 or Windows Server 2008 R2.

For more information about this you can check KB2517329

Cisco UCS Hyper-V Cluster – Add Nodes to Hyper-V Cluster – Part 7

Adding Hyper-V Hosts to the Failover Cluster is pretty simple. And I will keep this part very short.

  1. First Setup your Hyper-V Hosts and configure the them (network, storage, etc)
  2. Now open the Failover Cluster Manager and choose “Add Node…”
    Add Failover Cluster Node
  3. Select the new Hosts
    Add Failover Cluster Node
  4. After you check you have added the right Hyper-V Hosts click next
    Add Failover Cluster Node
  5. Now the hosts are added to the Failover Cluster
    Add Failover Cluster Node

Now you can add or move virtual machines to the new Cluster Nodes.

Cisco UCS Hyper-V Cluster – Create Hyper-V Cluster – Part 6

After you have installed your Hyper-V Nodes and enabled the Failover Cluster Feature you can now start two create the Hyper-V Server 2008 R2 Cluster.

This is not really difficult but you have two know about some important things.

  • The Failover Cluster Configuration Wizard will validate your Cluster configuration. If you have created a Hyper-V 2008 R2 Image with the latest drivers from Cisco your totally fine in the drivers check.
  • All SAN Disks have to be offline on the Hyper-V Hosts. Otherwise the Configuration Wizard can not validate the Cluster Disks.
  • As usual you need a clean Active Directory and a clean and working DNS Zone.
  • If you use Hyper-V Server Hosts you will get a warning in the network configuration because the standard firewall rule will not allow ICMP traffic (ping) to answer.
  • On the Cisco UCS you will also get a Warning which says your Cluster Network is maybe not redundant. This is because we added only one Management Network adapter for the Hyper-V hosts and with normal Hardware this is a single point of failure. But with on the UCS Hardware the Network adapter is virtual and the network connections are redundant in the background, so we can ignore this.

Now basically the setup of the Hyper-V Cluster is the same on the Cisco UCS as on every other Hardware.

  1. First start the Failover Cluster Manager, if you use the Hyper-V 2008 R2 Core Server you need the Failover Cluster Manager console on another Windows Server 2008 R2 or Windows 7 client. If you need this on a Windows 7 computer you can download the Remote Server Administration Tools for Windows 7.
  2. Now you first click validate the Configuration or on Create a Failover Cluster (this will also validate the Cluster Configuration)
    Microsoft Hyper-V 2008 R2 Failover Cluster
  3. Select your Hyper-V Host Servers
    Microsoft Hyper-V 2008 R2 Failover Cluster
  4. Run all tests
    Microsoft Hyper-V 2008 R2 Failover Cluster
  5. Now the Failover Cluster Configuration Wizard will validate your configuration.
    Microsoft Hyper-V 2008 R2 Failover Cluster
  6. After the validation you will get a report about the Configuration. Remember you with the Cisco UCS Solution and Hyper-V Server 2008 R2 you can get two warnings. The first about ICMP and the second because you have only one network adapter.
    Microsoft Hyper-V 2008 R2 Failover Cluster
  7. If you get no other warnings or errors you can now start to create the Failover Cluster. Add your Hyper-V Hosts.
    Microsoft Hyper-V 2008 R2 Failover Cluster
  8. Create the Access Point for the Failover Cluster (Clustername and IP Address)
    Microsoft Hyper-V 2008 R2 Failover Cluster
  9. Confirm the Configuration
    Microsoft Hyper-V 2008 R2 Failover Cluster
  10. And now the Wizard will create your Hyper-V Failover Cluster
    Microsoft Hyper-V 2008 R2 Failover Cluster
  11. After you have done this you simply have to add a Quorum Disk and activate Cluster Shared Volumes

 

Cisco UCS Hyper-V Cluster – Configure Hyper-V Networks – Part 5

This How-To shows you how you configure the (Virtual) Network Adapters of the Hyper-V Servers. This is not really heavy, but to complete the UCS Hyper-V Guide I post this. If you use Microsoft System Center Virtual Machine Manager 2008 R2 you will find later a post about doing this in SCVMM 2008 R2.

  1. Connect to the Hyper-V Server with the Hyper-V Manager Console
    Hyper-V Manager
  2. Now you can configure the Networks under Virtual Network Manager on each Hyper-V Host.
    Hyper-V Virtual Network ManagerWhat we did is, we added 7 (Virtual) Network adapters to the UCS Bladenodes in the UCS Manager. We added the same on Configuration on the Blades which are using VMware ESXi and on the Blades with Microsoft Hyper-V and thats why we have a Network called vMotion on the Hyper-V Servers. We use the vMotion network adpater for the private Failover Cluster Heartbeat.

Basically we have the following Networks:

UCS Blade Server Networks

  • 1. Network adapter is the Hyper-V Management Network dedicated to the Hyper-V Node
  • 2. Network adpater for Hyper-V Cluster Live Migration
  • 3. Network adapter for private Failover Cluster Heartbeat
  • 4. Network adapter External Network, is used for our main external Network
  • 5. Network adapter Internal Network, is used for our internal Management Network for Servers
  • 6. and 7. Network adapters are used for VLAN Trunks

To get the best performance we don’t share any Network Adapter with the Hyper-V Host and a Virtual Network.