The Microsoft support policy for server clusters, the Hardware Compatibility List, and the Windows Server Catalog

A typical HCL or WSC-compliant cluster contains the following components:

• Two or more servers that are referred to as the cluster nodes.
• A mass-storage controller and drives in each node from which the operating system boots.
• A separate mass-storage controller (SCSI or Fibre) that is independent of the controller that is used to boot the operating system.
• The shared disk. This is typically an external cabinet to which the mass-storage controller is attached that is connected to all nodes in the cluster.
• Two network adapters in each cluster node.

Note that adaptations of server clusters from OEMs may contain variations on these components, may not contain all of the components, or may contain more of the components that are listed. The key, however, to Microsoft support is that the entire cluster is listed on the HCL or WSC in the Cluster category.



However, there may still be configuration issues with an HCL-compliant or a WSC-compliant cluster. Therefore, we recommend that you run the Microsoft Cluster Configuration Validation Wizard to determine any potential configuration problems before you configure a Windows 2000 server cluster or a Microsoft Windows Server 2003 server cluster. To download the Microsoft Cluster Configuration Validation Wizard, visit the following Microsoft Web site:When you configure a Windows NT 4.0-based server cluster, the cluster installation routine displays the following agreement:
The Cluster Setup program in Windows 2000 displays the following agreement:

Windows Server 2003 does not display a dialog box that prompts for acceptance of the HCL or WSC agreement during the cluster configuration. The requirements for the Cluster service do state that an HCL cluster is required.

SAN, Geographically dispersed clusters, and Windows 2000 Datacenter Server

If multiple clusters are attached to the same storage area network (SAN) device, extra steps are required. The nodes in the cluster must be certified with the particular SAN device. The SAN device itself must be certified as Multi-Cluster capable and listed in the Cluster/Multi-Cluster Device category.

If the cluster solution includes geographically dispersed cluster nodes, additional items such as network latency and shared disk support must be verified. The entire cluster solution must be listed in the Cluster/Geographic category. See the following Microsoft Knowledge Base article for details:

280743 Windows clustering and geographically separate sites

Server clusters that run on Windows 2000 Datacenter Server have additional requirements for support. There may be certain contractual limitations to which you must adhere. See the following Microsoft Knowledge Base article for details:

265173 The Datacenter program and Windows 2000 Datacenter Server product

Clusters that are not listed on the HCL in the cluster category

If your cluster solution is not listed on the HCL or WSC in the Cluster category, the cluster solution is considered unsupported. However, PSS will offer troubleshooting tips if the customer requests them. PSS does not guarantee that a resolution will be found for non-HCL or non-WSC Clusters.

Steps for support

1. Before any troubleshooting begins, the customer must contact the OEM to discuss whether the particular cluster implementation is supportable. The OEM can best answer configuration and supportability questions for the cluster hardware.
   
2. Upon agreement that no solution is guaranteed, and that no incident refund will be given, PSS troubleshoots the issue. Microsoft does not guarantee a solution with non-HCL or non-WSC clusters. If no resolution is found, the incident is not refunded.
   
   If it is not agreed that a solution is not guaranteed, PSS will not troubleshoot the issue and will refund the incident.
3. Standard troubleshooting processes will be used to isolate the server cluster issue. Some typical troubleshooting methods that will be used by PSS include:
   • The Microsoft Knowledge Base is available to customers through Microsoft TechNet and the following Microsoft Web site:
     http://support.microsoft.com/support/
   • Determine whether the problem can be replicated on supported clusters (where possible).
   Note If the cluster is not certified, there is no hotfix support available. PSS will be unable to determine whether the problem is caused by a hardware incompatibility or undesirable software behavior.
   
4. If there is no solution to the problem, PSS may recommend some constructive alternatives, including:
   • Having the customer reproduce the problem on a cluster that is on the Cluster HCL or WSC.
   • Using a cluster solution that is on the Cluster HCL or WSC.
   • Having the customer work with the OEM to get the cluster on the Cluster HCL or WSC.
   • Working with the OEM for a solution.

The hardware specifications for server clusters are extremely stringent. The Cluster HCL and WSC contains a list of known acceptable cluster configurations that have been tested. You can waste a lot of time by attempting to troubleshoot perceived server cluster issues that are being caused by the cluster hardware that you are using.


Some examples of hardware incompatibilities that can cause cluster problems include:

• The cluster solution does not properly isolate the shared disk and HBAs from other devices on the shared bus.
• The SCSI controller does not support operating in a multiple-initiator environment.
• The HBA does not properly handle reservations or release or renew a device on the shared bus.
• The caching mechanism on the controller is incompatible with the cluster configuration.
• The network adapters for intra-cluster communication have too high a latency.
• The RAID controller does not properly replicate configuration information between controllers.
• The PCI bus is incorrectly configured and has incorrect adapters in the wrong bus (primary, secondary, tertiary, and so on).
• The controllers are incompatible the "Physical Disk Resource" type.
• The SCSI controller does not deploy proper termination.

This list does not include all the issues that can cause problems with a server cluster. None of these issues can be detected by PSS. All of these issues would typically be discovered if the complete cluster solution were tested for Cluster HCL or WSC compatibility.



If a complete cluster configuration is listed for an earlier operating system but is not listed for the newer operating system that you are using, support as documented in this section will be followed.