Microsoft support for server clusters with third-party system components (814607)

SUMMARY

Enterprise solutions are becoming more and more complex as companies develop different hardware technologies and as business requires such programs as mandatory antivirus software or quota management software. Microsoft does not provide many of these components. Typically, third-party software or hardware manufacturers provide them.

This could lead to confusion about whether a server cluster configuration is supported, the software components must appear on the Microsoft Windows Server Catalog, the software must be logoed, and other areas. This article discusses our support of server clusters that run third-party components, and what must be part of a qualified cluster solution. In particular, this article discusses the following software types:

• Network adaptor teaming
• Storage drivers (including SCSIPort/Miniport, StorPort/Miniport and Full port drivers)
• Multi-path storage drivers
• Storage management (such as volume managers and volume replication)
• Antivirus software
• Open-file backup agents
• Quota management software

Note Programs that include Microsoft programs such as Exchange 2000 Server or SQL Server 2000, have additional support requirements. This article discusses support from a Windows and a server cluster perspective only. This information is specific to Windows 2000 Advanced Server and Windows Server 2003 Enterprise Edition. The Datacenter program has much stricter requirements about what you can install on a Datacenter server for both Windows 2000 and Windows Server 2003.

back to the top

Server Cluster Qualification

Microsoft supports only qualified server cluster configurations. After Microsoft qualifies a cluster configuration and lists it on the Windows Server Catalog, Microsoft Product Support Services (PSS) provides support for the configuration based on the Service License Agreement and the support contract between Microsoft and the end-customer. All the components of a server cluster configuration must be appropriately logoed before Microsoft qualifies a cluster. For example, the servers, HBA and storage must be logoed and the complete cluster configuration must appear on the Windows Server Catalog in the cluster solutions hardware category.

Note Cluster configurations for both Windows 2000 and Windows Server 2003 appear on the Windows Server Catalog list.

back to the top

Qualified Server Cluster Configurations for Windows 2000

Microsoft supports a Windows 2000 cluster configuration only if that configuration appears on the Windows Catalog or HCL. Older Windows 2000 configurations only appear on the HCL. For more information about the support policy, click the following article number to view the article in the Microsoft Knowledge Base:

back to the top

Qualified Server Cluster Configurations for Windows Server 2003

The support requirements that this article previously discussed apply to Windows Server 2003. However, Microsoft also supports qualifications that are submitted through a new qualification process that is known as the Enterprise Qualification Process (EQP). In the EQP, the details of the specific driver versions, firmware and other details do not appear on the Windows Server Catalog. Instead, the Windows Server Catalog has a reference to a vendor Web site that contains the specific combinations of products product firmware and device driver version information that the vendor supports. Windows Server 2003 qualification does not qualify the Cluster service for Windows Server 2003 x64 qualification or IA-64 qualification. Separate qualifications for these operating system SKUs are required.

back to the top

Changes to Server Cluster Configuration that require retest by vendor

The Windows Server Catalog shows the components that make up the cluster and the qualified versions of the drivers and the firmware that the configuration requires.

Note A logoed configuration for a given operating system version is independent of service packs or hotfixesthat are applied to the system. However, there have been some compatibility issues between service packs and some vendor software components. Therefore, if the Service Pack or hotfix contains storage infrastructure drivers, the vendor should be consulted to determine their support for the service pack.

back to the top

Enterprise Qualification Program

This section applies only to partners in the enterprise qualification program (EQP). The EQP enables a partner to submit certain cluster configurations for listing in the Windows Server Catalog without having to provide the accompanying test logs.

Note Test logs are required for those cluster configurations that are specified in the clustering section of the EQP agreement.

If a company leaves or is disqualified from the EQP, that company can use the cluster solution (non-EQP) or cluster block models to submit the required test logs for the configurations for which they did not submit test logs when the company was an EQP partner. Any new cluster configurations a company makes after leaving the EQP must be submitted using the regular WHQL cluster submission and test process.
The following table lists the changes to a cluster configuration that qualify it as a new configuration for listing after the initial log requirement has been fulfilled.

Component	Change in configuration	Windows Catalog Submission
Operating system	Use a different operating system. For example, change from Windows Server 2003, Enterprise Edition (32-bit) to Windows Server 2003, Datacenter Edition (32-bit or 64-bit) or Windows Server 2003, Enterprise Edition (64-bit version).	Submission required; no test logs required
Server	Different server type.	Submission required; no test logs required
HBA (SCSI, iSCSI, Fibre Channel)	Adaptor change.	Submission required; no test logs required
New switch (Fibre Channel only)	Switch change.	Submission required; no test logs required
New hub (Fibre Channel only)	Hub change.	Submission required; no test logs required
New storage	Storage change.	Submission required; no test logs required
Single site -- Cluster size	Change in the number of supported nodes. The number of nodes in the new cluster is more than in the previous listing.	Submission required; no test logs required
Geo-cluster-cluster size	Change in the number of nodes supported. The number of nodes in the new cluster is more than in the previous listing.	Submission required; no test logs required
Multi-path software	Change from single path to multi-path.	Submission required; no test logs required
Shared Fabric environment (cluster deployed in a SAN)	Change from stand-alone cluster to shared fabric.	Submission required; no test logs required

During submission, Microsoft will not ask for test logs for any one of the cluster configuration changes listed in the previous table.

For any changes to the software, driver, firmware, or BIOS, the EQP partner must provide the appropriate versions for a given cluster configuration on their Web site as required by the EQP program.

Note The EQP partner must still follow the submission rules to qualify a device for a "Designed for Windows" logo, driver changes, and updates.
Note Vendors may provide updated versions of their software drivers and firmware that are qualified for server clusters. These changes may require a retest to provide an update Cluster server qualification based on the previous table. The vendors typically test a specific combination of driver version, host bus adaptor firmware, multi-path software version and controller version. Microsoft supports only complete vendor-qualified combinations. Microsoft does not support combinations that end users create.

back to the top

Not Enterprise Qualification Program (EQP)

To qualify for support, the configuration must have all the following items and the following changes require retest for non EQP submissions per the cluster specification:
For "Refresh Test to Achieve Logo," the cluster configuration must pass the following tests:

• Signed Driver Check test
• Validate N Node test

	No Retest Required	Full Retest	Refresh Test to Achieve Logo
Server	Changing the number of processors in a multiprocessor server. (Example 2->4 or4->2)
	Changing the number of processors in a multiprocessor server. (Example 2->1)	Changing from a single processor server to a multiprocessor server. That is, changing the number of processors in a single processor server. (Example 1->2)
	Changing the network card used for client access as long as this new card is in on the Windows Server Catalog.		New service pack, new hotfixes, or both, to server that has changes to the storage stack.
			Operating System service pack, hot fixes, or both, that include changes to storage stack.
	Changing the internal bus used to restart the system.	Changing the server that is used in the configuration. We determine a different server if there is a separate entry in the Windows Server Catalog for that server.
	Adding more memory to the server.
	Processor speed is equal to or less than the speed of the processors in the logoed server.		Processor speed is more than the speed of the processors in the logoed server.
	Changes to the BIOS settings or firmware version.
	Any changes to devices that are not on the cluster candidate component Windows Server Catalog, such as internal CD-ROM drives, tape drives, video cards, and so on.		Change or Add new drivers to any one of the nodes.
Storage / RAID Controller	Changing the RAID level that is used in HW RAID. Going from all RAID 1 sets to all RAID 10 sets for example does not make it a new configuration. When the SCSI devices are tested at Microsoft they will be configured and tested in all different configurations.	Changing or adding a SCSI HW RAID box of a different type than previously validated.	Updates to existing driver or filter driver or firmware.
	Adding another SCSI HW RAID box of the same type, as long as that RAID box is on a bus with a SCSI HBA previously in the configuration.		Additional storage I/O path(s).
	Adding more disks to a configuration.
Fibre or iSCSI Host Bus Adaptor	Adding another Fibre or iSCSI HBA of the same type, as long as the storage solution on that new bus is the same as the original tested configuration.		Changes to the HBA firmware or software driver.
	Changes to the support of the PCI bus speed. Some adaptor can support 33 MHz and 66 MHz.		Change to the BIOS.
Switches and Hubs (Fibre Channel only)	Changing the number of ports in a switch block.	Changing or adding a Fibre Channel switch or hub of a different type than previously validated.	Changing the Fibre Channel switch firmware to a version earlier than previously validated.
	Changing the Fibre Channel switch firmware (minor revision) to a version later than previously validated.		Changing the Fibre Channel switch firmware (major revision) to a version later than previously validated.
			Changes to a switch or the switch software.
SCSI Host Adaptor	Adding another SCSI HBA of the same type, as long as the storage solution on that new bus is the same as the original tested configuration.	Changing or adding a SCSI RAID adaptor of a different type than previously validated.
		Changing or adding a SCSI HBA of a different type than previously validated.
Geographically Dispersed Clusters		Change intersite interconnect technology (for example, T1 to ATM).	Change VLAN technology.
		Change quorum arbitration mechanism.	Update data mirroring s/w (if host-based s/w).
*** Communication Link		Different type of equipment.	Changes to firmware, software, or hardware components.

*** Communication Link is the hardware component that linked the two storages that are geographically separated. This equipment is usually found in a geo-cluster configuration.

back to the top

Multiple Clusters on a Single Storage Area Network

In large, complex storage area networks (SANs), several clusters may have access to a single storage controller. If more than one cluster (either multiple clusters or a single cluster and other non-clustered servers that run either Windows or any other operating system) uses storage on a single controller, the storage device must be multi-cluster device qualified and appear on the Storage/Raid Storage/Raid System/Multi-cluster category on the Windows Server Catalog. For more information, click the following article number to view the article in the Microsoft Knowledge Base:
back to the top

Starting From the Storage Area Network

Microsoft supports starting a Windows server from a device on a SAN (also known as "restart from SAN"). However, there are many configuration implications to this kind of deployment. Storage vendors have tested specific configurations that they support, and customers must work with the storage vendors to make sure that the customer's configuration meets the vendor's support requirements. The Windows Server Catalog configuration information for server clusters or for RAID System qualification does not list booting from the SAN, because you can set up any qualified cluster configuration to restart from the SAN in the limits of the Microsoft and storage vendor requirements. For additional information about the requirements of starting a Windows Server from a SAN, click the following article number to view the article in the Microsoft Knowledge Base:

back to the top

Cluster Components

Determine Who Supports the Components

Microsoft supports only qualified server cluster configurations. This means that Microsoft cannot debug or support software from third-party vendors, nor can Microsoft support and troubleshoot complex hardware issues alone. This section discusses how to determine who supports what.

A qualified configuration is a configuration that Microsoft supports (together with the vendors where appropriate) to provide root cause analysis and may escalate an issue found in the Windows operating system through the Windows escalation process. This process helps make sure that customers receive all the required hotfixes for the Windows operating system. Microsoft supports and troubleshoots the Windows operating system and the server cluster components to determine the root cause. To do this, you may have to turn off non-Microsoft products to isolate the issue to specific components or to reduce the number of variables. Microsoft intends to do this only if it does not affect the environment. For example, you can turn off quota management software, but you cannot turn off storage multi-path software unless you make infrastructure changes in the SAN. Or worse, if you turn off volume management software, programs may lose access to the data.

If the analysis determines that non-Microsoft components in the configuration are causing the issue, the customer and Microsoft together must work with the appropriate vendor. This occurs in the context of the agreements between the customer and the vendor. Microsoft does not provide a support path to vendors. Therefore, appropriate service agreements between the end customer and the third-party vendor must cover all the components.

back to the top

Storage Components

There are two kinds of storage components. A storage driver may be either a Full port driver or a SCSIPort/Miniport combination (in Windows Server 2003, Microsoft introduces a third concept of a Storport/Miniport combination). Different storage vendors provide different drivers. However, the type of driver affects the level of root cause analysis that Microsoft can provide for storage-related issues. A SCSIPort/miniport or a StorPort/miniport combination includes two components, the port driver that Microsoft provides and the miniport driver that the vendor provides. The port driver controls much of the storage operation. Therefore, Microsoft can provide complete troubleshooting and debugging for this. With a full port model, the vendor provides the complete stack. This means that Microsoft has no means to troubleshoot I/Os after they are passed to the driver.

In configurations that contain multiple storage fabrics for high availability, vendors provide multi-path software to control how the server sees the storage. These drivers are vendor-specific and sometimes are tied to specific hardware configurations. Again, Microsoft has no means to troubleshoot I/Os after they are passed to the multi-path driver.

If our analysis shows that storage issues are a potential cause of server cluster instability, especially in a SAN environment, Microsoft may require the storage vendor to validate the storage infrastructure before Microsoft continues with the root cause analysis. The validation makes sure that the vendor has qualified and supports the set of drivers and firmware combinations that the customer is using. (Later, this article discusses configuration issues.)

back to the top

Other Components in the Computer

The server cluster qualification covers the basic components of the computer such as the network cards and the basic storage infrastructure. It does not cover higher level functions. In particular, the server cluster qualification does not cover the following components:

• Network adaptor teaming
• Storage management (such as volume managers and volume replication)
• antivirus software
• Open-file backup agents
• Quota management software

back to the top

Network Adaptor Teaming

You can use network adaptor teaming (the ability to bind several network adaptors in a single logical network card) to provide a highly available network connection to a server. Microsoft supports server cluster configurations that use network adaptor teaming on the public interfaces. You must contact the vendor to verify that it has tested the network adaptor teaming drivers on a server cluster. Microsoft does not test network adaptor teaming configurations, nor does Microsoft provide troubleshooting and debugging for network adaptor teaming drivers.

If Microsoft determines that network issues may cause the instability issue in a server cluster, PSS may ask a customer to turn off the network adaptor teaming feature on the cluster. This reduces the number of components for root cause analysis. If Microsoft determines that network adaptor teaming is the root cause of the issue, Microsoft will ask the customer to contact the vendor for additional analysis.

For additional information about network configuration best practices that include how to use network adaptor teaming, visit the following Microsoft Web site:

back to the top

Components That Insert Drivers in the Storage Stack

Many components such as antivirus software, open-file backup agents, quota management software, and storage management software such as disk replication services or enhanced volume management services have filter drivers that are loaded in the storage stack. Microsoft does not provide a support path to vendors. Microsoft has a "Certified for Windows" logo program for software that tests basic functionality.

For additional information about the "Certified for Windows" logo program, visit the following Microsoft Web site: Microsoft uses this process to verify that software performs correctly in a cluster. Contact the vendor to make sure that the software that you want to deploy on a cluster is certified under the logo program for cluster operation. If storage issues are potentially causing instabilities in the cluster, PSS may prompt you to turn off the software during troubleshooting. You can do this with programs such as quota management software, antivirus software, or open-file agents. If Microsoft determines that the third-party product is causing the issue, Microsoft may ask the customer to contact the vendor for additional analysis.

For some third-party software components such as volume management products, you cannot turn off the software. In these cases, if Microsoft determines that the non-Microsoft components are the potential cause of the issue, the customer and Microsoft together must work with the appropriate vendor. This occurs in the context of the agreements between the customer and the vendor. Microsoft does not provide a support path to vendors. Therefore you must make sure that all the components are covered by appropriate service agreements between the end customer and the third-party vendors.

back to the top

Programs That Run on a Server Cluster

The support policies for the individual programs cover programs that run on a server cluster. Contact the program vendors to make sure that the vendor supports the program on a server cluster and to determine whether the vendor has requirements other than the requirements of the Windows operating system. We recommend that customers only run programs that are Certified for Windows and have run the cluster components of that logo program on a server cluster.

For more information about SQL Server 2000 support policies, click the following article number to view the article in the Microsoft Knowledge Base: back to the top

Determine Whether a Program is Certified for Windows

The cluster components of the Certified for Windows program are optional for the Windows 2000 Advanced Server and Windows Server 2003 Enterprise Edition logos. However, they are mandatory for Windows 2000 Datacenter Server and Windows Server 2003 Datacenter Edition. For programs logoed for Windows 2000, the Veritest Web site contains the details of the tests performed against logoed programs including whether the cluster components were run.

For additional information, visit the following Veritest Web site: Each product contains the Windows versions it is qualified for. For example, at the Veritest Web site, click Windows 2000 Datacenter Server or Windows 2003 Datacenter Edition in the Operating System box, and then click Search. The list contains companies that have programs logoed for Datacenter. Click to select one of the companies, and then click to open the certification report. Chapter 6 of the report shows whether the program works on a server cluster. The test results for Windows 2000 Advanced Server contain the following categories:

	Chapter 6 Cluster Service (Advanced Server and Datacenter only)
Yes	6.1 Applications must be able to install on 2 nodes for certification on Windows 2000 Advanced Server and 2,3, and 4 nodes on Windows 2000 Datacenter Server
Yes	6.2 Applications must support failover to all cluster nodes
Yes	6.3 Clients must survive the failure of the server program without crashing or affecting the stability of the system

The third-party products that this article discusses are manufactured by companies that are independent of Microsoft. Microsoft makes no warranty, implied or otherwise, regarding the performance or reliability of these products.
Microsoft provides third-party contact information to help you find technical support. This contact information may change without notice. Microsoft does not guarantee the accuracy of this third-party contact information.

In the Windows Server 2003 timeframe, you do not have to read through the test results to view that information in the Windows Catalog.

To view the Windows Server Catalog, visit the following Microsoft Web site:back to the top

Configuration Information

Specific components are qualified for use on a server cluster, but software components have many potential configuration options. The Hardware Compatibility List (HCL) does not cover these because they depend on the environment where you deploy the cluster. Microsoft provides the following information:

• A set of requirements about the configuration of any supported server cluster
• A set of best practices and guidelines that are independent of any specific vendor

For additional information about these requirements, best practices and guidelines, see the "References" section . Contact the vendor for the configuration rules and best practices for hardware or software deployment. The following sections are examples of configuration rules and best practices that the HCL does not list. However, you must follow these rules and best practices to successfully operate a server cluster.

back to the top

Network Configurations

To receive support from Microsoft, a cluster configuration must have at least two networks configured for intra-cluster communication. Typically this is a dedicated private network and a public network that is configured for both client access and intra-cluster communication. Depending on the deployment, a cluster may have other requirements.

For additional information about network best practices and requirements, visit the following Microsoft Web site: back to the top

Using the Quorum Disk for Programs

We recommend that the quorum disk is a dedicated disk and that programs do not use it to store data.

Isolating Clusters on a Fibre Channel Storage Area Network by Using Zoning or Masking

In a shared fibre channel SAN environment, you must isolate each of the clusters from any other clusters or servers. The arbitration algorithm uses SCSI commands that only affect disks associated with the cluster. We recommend that you zone fabrics at the switches to provide this isolation. However, some vendors have other techniques to isolate storage devices by using Logical Unit Number (LUN) masking or selective presentation mechanisms. Contact the storage vendors to determine the best practices and the correct way to isolate clusters. Issues in the storage fabric, especially with isolation, may corrupt your data.

For additional information about storage configurations and best practices, visit the following Microsoft Web site: back to the top

Additional Links and References

Web Sites

back to the top

Whitepapers and Best Practices

Server Cluster Network Requirements and Best PracticesServer Cluster Security Best PracticesServer Cluster Configuration Best PracticesServer Cluster Backup and Recovery Best PracticesWindows Clustering: Storage Area NetworksGeographically Dispersed Clustersback to the top

Microsoft Knowledge Base Articles

For more information, click the following article numbers to view the articles in the Microsoft Knowledge Base:
For more information, click the following article number to view the article in the Microsoft Knowledge Base: back to the top