HP Netserver High Availability SPOFless MSCS Cluster Solution

ManualsBrandsHP ManualsServerHP NetServer LC 3

september 2000

NetServer DocumentationNetServerNetserver

Introduction

This document profiles efforts in evaluating high availability cluster solutions

managed by Microsoft

Cluster Server in Windows NT 4.0

Ente

rprise Edition

and Windows

2000 Advanced Server operating systems. The solutions

evaluated are two-

node clusters using redundant Fibre Channel Arbitrated

Loop (FCAL) technology for the server to storage interconnect and are

intended to be highly reliable and highly available configurations with

Single Point of Failure (NSPOF) also known in HP as SPOFless.

High availability clusters are designed to reduce risk and costs of downtime

due to hardware or software faults. This high level of system and appli

cation

availability is accomplished through the use of redundant and fault tolerant

technologies implemented on different levels of hardware and software.

For example, clustering software improves availability by allowing applications

and services to be tr

ansferred to another computer node if a failure event

occurs on one of the nodes.

Overview of the SPOFless System

Achieving Higher Availability through Component Redundancy

SPOFless ensures that a system will be highly available and accessible in

case of a

ny system component failure. Using redundant hardware and

software components eliminates single points of failure. SPOFless system

redundancy falls into three categories.

• Redundant Hardware Components:

system drives, RAID sets, external

storage controlle

rs, host bus adapters (HBA), network adapters,

multiprocessor servers, Fibre Channel (FC) hubs, switches, power

supplies, processor power modules, and fans.

• Redundant Software Components: Microsoft

Cluster Server, network

fault tolerance and adaptive load

balancing, HBA and controller drivers

using firmware that supports fail over, and software mirroring.

• Fault Tolerant Techniques:

hot swappable hard drives, FC controllers,

hot plug/replace/upgrade PCI adapters, cache mirroring, and battery

backup.

Summary of content (9 pages)

PAGE 1
white paper HP Netserver NetServer Documentation HP Netserver High Availability SPOFless MSCS Cluster Solution Introduction This document profiles efforts in evaluating high availability cluster solutions managed by Microsoft® Cluster Server in Windows NT 4.0® Enterprise Edition and Windows® 2000 Advanced Server operating systems.
PAGE 2
white paper HP NetServer Netserver Documentation Configuration Techniques One type of highly available (HA) solution is the cluster. Clusters consist of two or more servers that share networking and storage. An example of a two-node cluster appears below. In the example, there are many single points of failure, assuming there are no redundant components in the system. For instance, if the shared storage in the cluster fails, the rest of the cluster fails as well.
PAGE 3
white paper HP NetServer Netserver Documentation Redundant Fibre Channel Array Controllers SPOFless systems use redundant HP Fibre Channel array controllers. When a pair of these FC array controllers is installed into an RS/12FC storage cabinet, they operate on two redundant Fibre Channel loops. The redundant FC loops allow a failover to occur at the storage level in the event of a controller failure. After failover, the remaining controller supports both FC loops at a reduced performance capacity.
PAGE 4
white paper HP NetServer Netserver Documentation FC Array Controller Firmware The HP FC array controllers must have the firmware version 5.49-00 installed. The firmware version can be verified by using the FC Array Assistant. Since the RS/12FC firmware is independent of the operating system installed on the Netservers, the same firmware is used for both Windows® 2000, and Windows NT®. Network Interface Cards The driver version Hptxnt5.sys version 4.02.27.
PAGE 5
white paper NetServer Netserver HP Documentation Functional Diagram of the Direct Connect Cluster (SPOFless) Public Network Cluster Server #1 Cluster Server #2 Private Network FC HBA External Shared Storage Port 0 Port 1 SCSI #1 SCSI #2 SCSI #1 Copper Fibre Active Failure Causes Application Failover Port 0 SCSI #2 Most of the procedure for installing a SPOFless cluster is identical to the installation of a SPOF cluster solution.
PAGE 6
white paper HP NetServer Netserver Documentation FC Array Assistant Controller Information Screen FC Array Assistant Manual Configuration Screen – Mirrored LUN Mapping 6
PAGE 7
white paper HP NetServer Netserver Documentation together, all of the parameters should be checked using the FC Array Assistant utility. For complete information on how to use FC Array Assistant refer to the References section at the end of this document. The second issue has to do with the mapping of LUNs. When the disk arrays are configured, the LUNs are not automatically mapped in a mirrored configuration.
PAGE 8
white paper HP NetServer Netserver Documentation Casualty Procedures Failed FC Array Controller In the event of an FC controller failure, the following steps should be taken: • Obtain a new controller and upgrade the firmware to the same version of firmware as the working FC controller. If the firmware between the two controllers does not match, the storage functionality will be disrupted and the cluster will fail. Perform the upgrade on a separate server with an RS/12FC. Do not use a clustered RS/12FC.
PAGE 9
white paper NetServer Netserver HP Documentation References Functional Diagram of the Dual HBA Cluster ·HP Netserver Microsoft® Cluster Configuration Guides: http://netserver.hp.com/products/high_availability/ms _supported_config.