Network-attached storage and server cluster support

In Windows 2000 Server and the Windows Server 2003 family of products, the quorum disk must be a shared, block-level disk device that is accessible from all members of the server cluster. For a server cluster to be a supported configuration, the complete cluster configuration must appear on the Cluster Hardware Compatibility List (HCL). For more information, click the following article number to view the article in the Microsoft Knowledge Base: For Windows 2000, a prerequisite for qualification is that all components of the cluster solution are device qualified and cluster device qualified. The whole solution is listed on the Cluster HCL. For Windows Server 2003, a storage block qualification process is the prerequisite. Only device-qualified and cluster-device-qualified, block-level disk storage solutions can be used as disk resources in a server cluster.

You can store application data for cluster-aware applications on block-level disk devices that are managed by the server cluster by using a physical disk resource or a third-party resource monitor. The data can be accessed by means of drive letters or mount points, or it can be accessed on network-attached storage devices. These devices can be accessed by using Universal Naming Convention (UNC) paths of the form \\servername\sharename. For more information about the specific application requirements and support policies, see the "References" section later in this article.

Windows Server 2003 provides an alternative, optional quorum mechanism that is known as majority node set. By using this quorum mechanism, you can build cluster configurations without using a shared, block-level disk device as the quorum disk. In a majority node set server cluster, the quorum data is mirrored on the system disks of all the nodes in the cluster instead of having a single copy on the quorum disk. The Cluster service makes sure that the data is always kept consistent across the different mirrored copies.

In these cluster configurations, the Cluster service itself does not require access to a shared, block-level disk device for a quorum resource. Applications can continue to use either of the following:

• Block-level disk devices that are managed by the server cluster by using disk resources and that are accessed by means of drive letters or mount points
• Network-attached storage devices that are accessed by using UNC path names

As with all server clusters, these configurations must be on the Cluster HCL. For more information about the majority node set quorum mechanism, see Online Help and Support for Windows Server 2003.

Server clusters provide a highly available infrastructure for applications so that if one node fails, or if an application on one node fails, the application and its data can be moved to another node (a process that is known as failover). Application data is moved by changing ownership of the physical disk from one node to another. To make sure that only a single node can access or modify data on any particular disk at any point in time, the Cluster service implements an ownership-and-protection mechanism that is known as arbitration. As part of the arbitration algorithm, the Cluster service uses small computer system interface (SCSI) commands, in particular the scsi reserve and scsi release commands, and Logical Unit Number (LUN), Target, or bus reset. For more information, click the following article number to view the article in the Microsoft Knowledge Base: The cluster storage device qualification tests and the server cluster qualification tests make sure that a storage subsystem correctly implements the commands that are used for arbitration. The tests also make sure that, under heavy loads and multiple failure conditions, the cluster can always guarantee that only one node can access a particular device.

Microsoft provides tests for industry-standard storage solutions only. Currently, the tests cover parallel-connected SCSI disk storage solutions and fiber-channel-connected disk storage solutions that use SCSI protocols over fiber channel. Vendor-specific solutions that use proprietary storage interconnects or vendor proprietary protocols that transmit data over industry-standard interconnects (such as Ethernet) are not supported by Microsoft and cannot be qualified as part of a server cluster solution. As new storage technologies become industry standards, Microsoft will expand its qualification program to cover other solutions.

Additionally, the server cluster qualification tests evaluate only the basic functionality of the storage unit. If you enable vendor-specific features that are not integrated with the base Windows operating system, the storage vendor must address any storage-related issues with the cluster solution. Vendor-specific features include, for example, snapshots and disaster recovery solutions. Microsoft may require that the unsupported features be disabled before validating or troubleshooting the configuration.

Note Snapshot capabilities that are integrated with the virtual snapshot service (VSS) as part of Windows Server 2003 are fully supported in a server cluster.

The cluster qualification covers only those disk storage devices that are managed by the cluster (that is, devices that are failed over by the Cluster service where there is a specific server cluster resource type that is associated with the storage device). Applications can store their data outside the cluster on storage devices that are not managed by the cluster. Network-attached storage devices can be used by applications and do not have to be part of the Cluster service qualification. However, some applications have specific requirements about the use of network-attached storage. For more information about Microsoft products and their support requirements, see the "References" section later in this article.

For all cluster configurations, the following rules apply:

• Any shared, block-level disk devices that are managed by the cluster (that is, devices that are failed over by the Cluster service and have a resource type that is associated with them) must be qualified as part of the cluster configuration, and the complete configuration must appear on the Cluster HCL.
• Failover applications that run on a server cluster can store data on a storage device that is accessed by means of a UNC path. However, different applications have different restrictions, requirements, and support policies about network-attached storage devices. Make sure that your application supports the storage of its data on a network-attached storage device.

For more information about our support for iSCSI, visit the following Microsoft Web site:For a list of frequently asked questions about iSCSI, visit the following Microsoft Web site: For more information about server cluster support requirements, click the following article number to view the article in the Microsoft Knowledge Base: For more information about the server cluster arbitration mechanism, click the following article number to view the article in the Microsoft Knowledge Base: For more information about the specific storage requirements and support statements for other Microsoft products, click the following article numbers to view the articles in the Microsoft Knowledge Base: