MCSE 70-223 Study Guide

This exam tests your ability to install, configure, and administer clustering services using Microsoft Windows 2000 Advanced Server. Clustering is typically deployed, as according to Microsoft, in medium to very large computing environments that use the Windows 2000 Server operating system with the following characteristics: 

• Supported users range from 200-26,000+
• Physical locations range from 5-150+

• Typical network services and applications include file and print, database, messaging, proxy server or firewall, network and storage hardware, desktop management, Active Directory™ Services, TCP/IP connectivity, RAID, SCSI and Fibre Channel storage, and Web hosting.

• Connectivity needs include connecting individual offices and users at remote locations to the corporate network and connecting corporate networks to the Internet.

When you pass this exam, you can achieve Microsoft Certified Professional status. You also earn credit toward the following certifications:

• Elective credit toward Microsoft Certified Systems Engineer on Microsoft Windows 2000 certification
• Elective credit toward Microsoft Certified Systems Engineer on Microsoft Windows NT 4.0 certification

A lot of the information in this study guide is useful for the 70-226 exam as well. You should study both before attempting the exam.

What is a server "cluster"?

"A server cluster is a group of independent servers managed as a single system for higher availability, manageability, and scalability."

In the NT4 Era, cluster was offered in the "Wolfpack" as Microsoft Cluster Server (MSCS). With Windows 2000 Advanced Server and Datacenter server, cluster support becomes built-in. 

Clustering, as according to webopedia.com, is a method of:

"Connecting two or more computers together in such a way that they behave like a single computer. Clustering is used for parallel processing, load balancing and fault tolerance… Clustering is a popular strategy for implementing parallel processing applications because it enables companies to leverage the investment already made in PCs and workstations. In addition, it's relatively easy to add new CPUs simply by adding a new PC to the network."

Hands-on experience on cluster server is essential. The absolute minimal hardware needed for a minimum working configuration requires 2 servers mutually connected using 100BaseT connection with NO hub in between (you need a connection that is as fast as possible) running Windows 2000 Advanced Server or Datacenter Server. The evaluation edition of Windows 2000 Advanced Center is available for download at www.windows2000.com. More advanced setup involves the use of shared storage device. This device, however, is way too expensive for the purpose of exam preparation...

Cluster Basics

Requirements:

• Two servers connected by a network
• A method for each server to access the other's disk data – often achieved via a shared disk
• Special cluster software service

Cluster Availability:

• Automatically detect the failure of an application or server
• Quickly restart the application on a surviving server.
• A momentary pause in service exists

Cluster Manageability:

• You can quickly inspect the status of all cluster resources
• You may move workload around onto different servers within the cluster for the purpose of manual load balancing and "rolling updates"

Cluster Scalability:

• "Cluster-aware" applications can enjoy dynamic load balancing
• "Cluster-aware" applications can scale across multiple servers within a cluster
• Non-cluster aware applications can run, but cannot enjoy the benefits 

Surviving server get access to failed server's disk data via the following ways:

• Shared disks

• Mirrored disks
• "Shared nothing"

Shared disks

• Every server can access every disk.
• Requires specially modified applications.
• Inherent limits on scalability due to DLM contention

Mirrored disks

• Each server has its own disks
• Run software that "mirrors" every write from one server to a copy of the data on another server.
• Great for keeping data at a disaster recovery site in synch

Shared nothing

• Each server owns its own disk resources

• Servers share "nothing" at any point in time
• When a server fails, disk ownership is transferred from one server to another.

• Provides high level of availability

• Provides higher scalability

• Works with standard applications
• used by Windows 2000 Cluster

Cluster Service VS NLB

• The two Advanced Server clustering technologies are:
  • Cluster service
  • Network Load Balancing (NLB)
  • They can be used independently
  • They can be used in combination
• A server cluster:
  • is a set of independent node servers using connected storage devices
  • everything is managed together
  • Windows 2000 Advanced Server supports two-node clusters.
• Clustered servers:
  • are physically connected by cables
  • are programmatically connected by the cluster service software
  • are not necessarily be of the same size / same configuration.

Cluster Service

• The cluster appears to be a single system to the clients and to the applications.
• Should one server fail, the failover process will automatically shift its workload to another server in the cluster.
• Limited number of nodes
• This is ideal for high availability
• This does not provide load balancing

Load Balancing

• NLB supports clusters for front-end applications and services
• No need for shared storage
• Ideal for:
  • Internet and intranet sites
  • Web-based applications
  • Media streaming
  • Terminal Services
• NLB modes:
  • Single network adapter in unicast mode:

• works with all routers

• ordinary network communication among hosts isn’t possible
• network performance may be compromised.
  • Multiple network adapters in unicast mode:
• works with all routers
• ordinary network communication among hosts is possible
• network performance may be enhanced
• at least two network adapters are required.
  • Single network adapter in multicast mode:
• Only one network adapter is required
• ordinary network communication among hosts is possible
• network performance may suffer
• some routers may not support multicast MAC address

  • Multiple network adapters in multicast mode:

• Performance may be enhanced
• ordinary network communication among hosts is possible
• at least two network adapters are required
• some routers may not support multicast MAC address

Cluster Service + NLB

This kind of cooperation is possible with an n-tier architecture.

According to webopedia.com,

"N-tier application architecture provides a model for developers to create a flexible and reusable application. By breaking up an application into tiers, developers only have to modify or add a specific layer, rather than have to rewrite the enitire application over, if they decide to change technologies or scale up. In the term "N-tier," "N" implies any number -- like 2-tier, or 4-tier; basically, any number of distinct tiers used in your architecture."

You can, for example, use NLB for the front-end Web server farm, and use clustering for the back-end databases. In fact, it is technically difficult to load balance a database. Keeping the database in sync in a load sharing set up could be highly difficult:

This information is extracted from Microsoft web site.

Cluster hardware and drivers

• Check that your hardware, including your cluster disks and other cluster hardware, is compatible with Windows 2000 Advanced Server.
• Hardware Compatibility List is available on the Windows 2000 CD-ROM’s Support folder Hcl.txt file. It is also available from the web:

HCL location: http://www.microsoft.com/windows2000/server/howtobuy/upgrading/compat/search/devices.asp

According to Microsoft,

 "HCL is a Web-based searchable database which is updated as additional hardware is tested and approved."

• Check that you have the drivers supplied by your hardware manufacturer.
• Cluster hardware installation is not something easy. You must review the manufacturer's instructions carefully before proceeding.
• Cluster hardware for sure involves SCSI. Make sure you do the termination and the SCSI ID configuration right. In many times error comes from mis-configured SCSI items.

Installation Guideline

• Install the Cluster service one server at a time.
• Make sure that Windows 2000 Advanced/Datacenter Server and the Cluster service are installed and running on one server before starting the operating system and installing the Cluster service on a second server. This prevents the cluster disks from becoming corrupted.

Cluster models

• Model 1: Single node server cluster configuration
  • No failover.
  • Simply organize resources on a server for administrative convenience and for the convenience of your client
    More information:
    http://www.microsoft.com/WINDOWS2000/en/datacenter/help/sag_mscs2planning_7.htm
• Model 2: Dedicated standby role
  • provides maximum availability
  • provides maximum performance
  • expensive hardware
  • not maximum use of hardware
  • multiple primary nodes support all clients
  • companion/standby node is a dedicated server that is ready to be used whenever a failover occurs
    More information:
    http://www.microsoft.com/WINDOWS2000/en/datacenter/help/sag_mscs2planning_8.htm
• Model 3: High-availability configuration
  • provides reliability and acceptable performance when only one node is online
  • provides high availability and high performance when all nodes are online
  • allows maximum use of hardware resources
  • each node has its own set of resources available to the network in the form of virtual servers
  • virtual servers can be detected and accessed by clients
  • all nodes are active
  • no standby node
    More information:
    http://www.microsoft.com/WINDOWS2000/en/datacenter/help/sag_mscs2planning_9.htm 

Network adapters on the cluster nodes

• Each cluster node needs at least one network adapter.
• In any case, having multiple adaptors avoid single point of failure.
• Only node-to-node communication:
  • you connect the network adapter to a private network used exclusively within the cluster.
  • server has one or more additional adapters to carry other communication.
• Only client-to-cluster communication:
  • you will connect the network adapter to the appropriate network in your organization
  • server has one or more additional adapters to carry other communication.
• Both node-to-node and client-to-cluster communication:
  • multiple network adapters per node to carry both kinds of communication and provide backup for each others
• Communication unrelated to the cluster:
  • provides services unrelated to the cluster via an independent adapter

Cluster IP addressing

• Use a static IP address for the cluster itself.
• You can provide IP addressing through DHCP for the nodes.
• HOWEVER, the DHCP service itself may become a single point of failure
• You can assign each node’s network adapter with static IP address
• In any case, ensure that the addresses for each linked pair of network adapters are on the same subnet.
• If an interconnect connects only the cluster nodes and does not support any other network clients, you can assign it a private address:
  10.0.0.0 through 10.255.255.255 (Class A)
  172.16.0.0 through 172.31.255.255 (Class B)
  192.168.0.0 through 192.168.255.255 (Class C)

Cluster naming

• The cluster name is the name you use for connections to the cluster.
• The cluster name must be different from the domain name
• The cluster name must be different from all computer names on the domain
• The cluster name must be different from other cluster names on the domain

Computer accounts for the cluster nodes

• The nodes of the cluster MUST have computer accounts.
• These accounts MUST BE in the same domain.

User account for the Cluster service

• The Cluster service needs a name and password to run.
• You need to have the user name and password ready during cluster installation.
• To configure the Cluster service, you must first log on with this account. That is why you must create this user account in advance.
• This account must have local administrative privileges on the cluster nodes.
• This account should not be allowed to expire.

Quorum resource and cluster database

• Quorum resource:
  • maintains the configuration data necessary for recovery of the cluster
  • stored in the form of recovery logs
  • includes details of all of the changes that have been applied to the cluster database
  • Windows 2000 uses it to ensure that only one set of active and communicating nodes can operate as a cluster
  • Keep in mind that a node can join a cluster or remain in an existing cluster only if it can communicate with the node that controls the quorum resource.
• Cluster database:

  • when a node joins a cluster, the Cluster service will update the node's private copy of the cluster database.

  • when a node joins an existing cluster, the Cluster service will retrieve the data from the other active nodes.

Disk planning for the Cluster disks

• You need space for holding important cluster configuration information such as the information that makes up the quorum resource of the cluster. You need a minimum of 5MB for this purpose, although 500 MB is recommended by Microsoft.
  According to the technical documentation at Microsoft's web site, 
  
  "It is strongly recommended that you install the quorum resource on a separate disk, by itself, so that other applications cannot affect it. The disk should be at least 500 megabytes (MB) in size."
   
• The cluster disks should be configured as basic disks with all partitions formatted as NTFS.
• Do not use dynamic disks nor mount points on cluster disks.
• Do not use FAT / FAT32 on cluster disks.

Changing Quorum resource settings in the future

• You may change the quorum-resource location via the Cluster Administrator:
  1. In the console tree, click the name of the cluster.
  2. On the File menu, click Properties, and then click the Quorum tab.
  3. In Quorum resource, click the disk resource you want to use as the quorum resource for the cluster.
  4. In Partition, click the partition where you want the cluster specific data kept.
  5. In Root path, type the path to the folder on the partition.
• You may change the maximum quorum-log size via the Cluster Administrator:
  1. In the console tree, click the cluster.
  2. On the File menu, click Properties.
  3. Click the Quorum tab.
  4. In Reset quorum log at, type the maximum size that the quorum log file can be before the cluster resets it.
  5. Size is specified in kilobytes.

In the future, if you need to change the quorum disk designation, you better do it with all cluster nodes present, or when other cluster nodes attempt to join the existing cluster they might not be able to connect to the quorum disk.

In case you need to recover from a corrupted quorum log or quorum disk, go through the steps listed in http://www.microsoft.com/windows2000/en/datacenter/help/default.asp?url=/WINDOWS2000/en/datacenter/help/cluad_pr_70.htm

• Cluster service can manage:

  • Server cluster networks
  • Network interfaces
  • Nodes
  • Resource groups
  • Resources
• Resource groups:
  • logical collections of resources
  • made up of logically related resources, such as applications and their associated peripherals and data
• Resource:
  • any physical or logical component that can be brought online and taken offline
  • can be managed in a server cluster
  • can be owned by only one node at a time
  • when planning the resource groups for a cluster on your network,
  • you should list the dependencies for each resource
  • dependency list should include all resources that support the core resource.

Special Cautions mentioned by Microsoft

As quoted directly from Microsoft web site at http://www.microsoft.com/windows2000/en/advanced/help/default.asp?url=/WINDOWS2000/en/advanced/help/planning_cluster_installation.htm:

"When you first attach your cluster hardware (the shared bus and cluster disks), be sure to work only from the BIOS configuration screens on the cluster nodes (a node is a server in a cluster). The instructions from your manufacturer will describe whether the BIOS configuration screens display automatically or whether you must, after turning on the computer, press specific keys to access the BIOS. Follow the manufacturer's instructions for completing the BIOS Setup process. Remain in the BIOS, and do not allow the operating system to start, during this initial installation phase.

After the BIOS configuration is completed, start the operating system on one cluster node only, and carry out the installation of the Cluster service. Make sure that Windows 2000 Advanced Server and the Cluster service are installed and running on that node before starting the operating system on another node. If the operating system is started on multiple nodes before the Cluster service is running on one node, the cluster disks could be corrupted."

Installing the Cluster service

• You do this through Control Panel’s Add/Remove Programs
• In the Add/Remove Programs dialog box, click Add/Remove Windows Components. In the Windows Components Wizard dialog box, make sure that the Cluster Service check box is selected, and then click Next. Let the Windows Components wizard run.

After successful installation, you manage the cluster via the Cluster Administrator. A copy of Cluster Administrator is automatically installed on a cluster node when you install the Cluster service.

Cluster Administrator

• It allows you to:
  • manage cluster objects
  • establish groups
  • initiate failover
  • handle maintenance
  • monitor cluster activity
  • provide own resource types by writing extension
• To connect to a cluster with Cluster Administrator:
  • In Cluster Administrator, on the File menu click Open Connection.
  • In Cluster or server name, type the network name (or IP address) of the cluster or a node in the cluster.
• To configure applications to run on the cluster, you use the new Cluster Application wizard: http://www.microsoft.com/windows2000/en/advanced/help/default.asp?url=/WINDOWS2000/en/advanced/help/SAG_MSCS4managing_2.htm 

Running applications on NLB

• NLB can scale any application or service that uses TCP/IP
• Application must be designed to allow multiple instances to run simultaneously
• Don't use NLB to directly scale applications that independently update inter-client state data
• IIS is an ideal application to run on NLB
• SQL Server and Exchange Server should not be run on the NLB cluster because they independently update inter-client state data. You should use the Cluster service for them instead.

Hands-on experience on the Cluster Administrator UI is critical for passing the exam!

Reference Books

MCSE Clustering Using Advanced Server 2000 Exam Cram (Exam: 70-223)

Diana Bartley, Jarret W. Buse; Paperback