Planning for High Availability
Updated: April 29, 2011
Applies To: Virtual Machine Manager 2008, Virtual Machine Manager 2008 R2, Virtual Machine Manager 2008 R2 SP1
This topic explains how to plan for high availability of four components of System Center Virtual Machine Manager (VMM) 2008 and System Center Virtual Machine Manager 2008 R2—virtual machines, the VMM database, the VMM server, and library servers—and how to help ensure availability of the Web servers that host VMM Self-Service portals, for which high availability is not supported. High availability of a component ensures that the component remains functional, in some cases after a brief interruption, when the component or the computer running the component fails.
Planning for Highly Available Virtual Machines
VMM 2008 supports highly available virtual machines, also known as HAVMs, deployed on failover clusters created in Windows Server 2008 Enterprise Edition or Windows Server 2008 Datacenter Edition. VMM 2008 R2 supports the enhanced capabilities provided by failover clusters created in Windows Server 2008 R2 Enterprise Edition or Windows Server 2008 R2 Datacenter Edition. These failover clusters support as many as 16 cluster nodes in a single cluster.
For a detailed discussion of configuring and managing host clusters using VMM 2008, see Configuring Host Clusters in VMM to Support Highly Available Virtual Machines.
For information about the enhanced features offered by Windows Server 2008 R2 Hyper-V, see the white paper Windows Server 2008 R2 & Microsoft Hyper-V Server 2008 R2 - Hyper-V Live Migration Overview & Architecture (http://go.microsoft.com/fwlink/?LinkId=158868).
|To evaluate Windows Server 2008 R2, you can download the pre-release version from Windows Server 2008 R2 Release Candidate (http://go.microsoft.com/fwlink/?LinkId=146649). For information about availability of the released version of Windows Servers 2008 R2, see Introduction to Windows Server 2008 R2 (http://go.microsoft.com/fwlink/?LinkId=159732).|
Among the new Hyper-V features, VMM 2008 R2 supports the cluster shared volumes (CSV) feature of Windows Server 2008 R2. CSV enables multiple highly available virtual machines to share the same logical unit number (LUN) but still migrate independently without affecting other HAVMs that are sharing the same LUN. Because the HAVM can access the disk from a separate node, CSV also expedites migration of virtual machines within the cluster because there is no need to unmount and then mount the disk. CSV is not supported by VMM 2008. Unlike VMM 2008 R2, VMM 2008 does not support CSV. VMM 2008 supports only one highly available virtual machine per LUN. In VMM 2008, that is required in order to ensure that HAVMs can be migrated to different hosts without affecting the availability of other HAVMs.
VMM 2008 discovers highly available virtual machines, created outside VMM, that share a single LUN, but assigns the virtual machines an Unsupported Cluster Configuration status. To manage those virtual machines in VMM 2008, change the disk configurations outside VMM and then use the Repair action with the Ignore option to refresh each virtual machine’s status in VMM.
VMM 2008 R2 also supports the use of SAN transfers to migrate virtual machines and highly available virtual machines from one host cluster to another, into a host cluster from a non-clustered host, or out of a host cluster to a non-clustered host. When you migrate a virtual machine into a cluster from a non-clustered host by using a SAN transfer, VMM checks all nodes in the cluster to ensure that each node can see the LUN and automatically creates a cluster disk resource for the LUN. Even though VMM automatically configures the cluster disk resource, it does not validate it. You must use the Validate a Configuration Wizard in Failover Cluster Management to validate the newly created cluster disk resource. To migrate a virtual machine out of a cluster to a non-clustered host, the virtual machine must be on a dedicated LUN that is not using CSV.
|It is recommended that you not deploy virtual machines that are not highly available on your host clusters. Although you can do this by using Hyper-V (VMM does not allow it), the non-highly available virtual machines will consume resources that otherwise would be available to the HAVMs.|
|You can manage host clusters created in a VMware Infrastructure 3 environment by using VMM 2008. For more information, see Managing a VMware Infrastructure in VMM.|
Configuring Cluster Reserves
Depending on your needs, you can configure a cluster reserve for each host cluster that specifies the number of node failures a cluster must be able to sustain while still supporting all virtual machines deployed on the host cluster. If the cluster cannot withstand the specified number of node failures and still keep all of the virtual machines running, the cluster is placed in an Over-Committed state, and the clustered hosts receive a zero rating during virtual machine placement. The administrator can, during a manual placement, override the rating and place an HAVM on an over-committed cluster.
For example, if you specify a node failure reserve of 2 for an 8-node cluster, the rule is applied in the following ways:
If all 8 nodes of the cluster are functioning, the host cluster is marked Over-committed if any combination of 6 nodes (8-2) in the cluster lacks the capacity to accommodate existing virtual machines.
If only 5 nodes in the cluster are functioning, the cluster is marked Overcommitted if any combination of 3 (5-2) nodes in the cluster lacks the capacity to accommodate existing virtual machines.
|In VMM 2008 R2, placing a clustered host in maintenance mode can cause the cluster to enter an Overcommitted state. When a host is placed in maintenance mode, VMM either performs a live migration to move all virtual machines to a different host in the host cluster or places the virtual machines in a saved state. While a host is in maintenance mode, VMM blocks virtual machine creation on the host and excludes the host from host ratings during virtual machine placement so that no virtual machines will be migrated to it. For more information about maintenance mode, see Creating and Managing Highly Available Virtual Machines in VMM.|
Planning for a Highly Available VMM Database
To make the VMM database highly available, you can deploy the SQL Server instance on a failover cluster created in either Windows Server 2008 or Windows Server 2003.
To support a highly available VMM database, your VMM deployment must meet the following requirements:
An edition of Microsoft SQL Server 2005 or SQL Server 2008 that supports failover clustering is required.
The SQL Server instance must reside on a failover cluster created in Windows Server 2008 Enterprise Edition, Windows Server 2008 Datacenter Edition, Windows Server 2003 Enterprise Edition, or Windows Server 2003 Datacenter Edition.
The database server must be in an Active Directory domain that has a two-way trust relationship with the domain that contains the VMM server. Using a clustered database with VMM requires mutual authentication under Kerberos. To support this, the SQL Server instance must associate a Service Principal Name (SPN) with the account that SQL Server will be running on. For more information, see Registering Kerberos Service Principal Names by Using Http.sys (http://go.microsoft.com/fwlink/?LinkId=128396).
To help ensure successful database failovers, you might need to modify VMM's built-in retry mechanism for the VMM database. The retry mechanism has the following default settings:
Number of retries: 5
Retry interval: 2 seconds
If your SQL Server database resides on a cluster, the failover duration should be shorter than the product of the preceding retry values. If you need to change those values, you can create the following registry keys for the Windows registry on the VMM server. (The sample code shows the default values.) To open the Registry Editor, click Start, click Run, and then type regedit.
|Serious problems might occur if you modify the registry incorrectly by using Registry Editor or by using another method. These problems might require that you reinstall the operating system. Microsoft cannot guarantee that these problems can be solved. Modify the registry at your own risk.|
HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Microsoft System Center Virtual Machine Manager Server\Settings\SQL Value Name=DBRetryInterval Type=REG_DWORD Sample value=00000002 HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Microsoft System Center Virtual Machine Manager Server\Settings\SQL Value Name=DBRetryCount Type=REG_DWORD Sample value=00000005
Planning for a Highly Available VMM Server
Running the VMM server inside a highly available virtual machine is supported. Use Failover Cluster Manager in Windows Server 2008 or Windows Server 2008 R2 to create a failover cluster. Then use Hyper-V to create and manage the HAVM in which you want to run the VMM server. To avoid accidentally migrating, pausing, saving, or deleting the virtual machine that contains the VMM server, if you run the VMM server in a virtual machine, it is recommended you not use VMM to manage that virtual machine.
In the event of a failure of the VMM server, successful recovery depends on implementing a reliable backup plan for the VMM database. For information about recovery scenarios in VMM, see Backing Up and Restoring the VMM Database in VMM 2008 Help.
|Clustering of the VMM application, which enables failover of VMM itself, has not been tested and is not supported.|
Planning for Highly Available Library Servers
VMM 2008 supports using highly available file servers and shares hosted by a failover cluster created in Windows Server 2008 Enterprise Edition or Windows Server 2008 Datacenter Edition as highly available library servers and library shares. VMM 2008 is not aware of failover clusters created in Windows Server 2003. VMM 2008 R2 supports using highly available file servers and shares hosted by a failover cluster created in Windows Server 2008 Enterprise Edition or Windows Server 2008 Datacenter Edition, or in Windows Server 2008 R2 Enterprise Edition or Windows Server 2008 R2 Datacenter Edition, as highly available library servers and library shares.
For information about adding and managing highly available library servers, see Planning for the VMM Library.
Ensuring Availability of VMM Self-Service Portals
To ensure the availability of VMM Self-Service Portals for your self-service users, it is recommended that you maintain multiple portals on multiple Web servers. Virtual Machine Manager does not support Network Load Balancing (NLB) clusters in Windows Server 2008, which are required in order to distribute the network traffic among self-service users on multiple Web sites.