Windows Server 2008 R2 Failover Clustering and Network Load

Balancing

Kaliyan Selvaraj Technical Consultant, Dell India Pvt. Ltd.,

October 25th 2009

3

AgendaIntroduction to Clustering

What is clusterWhy use clusterCluster common terminologyWhat is High AvailabilityHardware / Software RequirementsCluster Validations / Cluster Creations.

Introduction to WNLBOverview of Network Load Balancing Clusters Configuring a Network Load Balancing Cluster Maintaining a Network Load Balancing Cluster

4

What Are Clusters?

A cluster is a group of computers and storage devices that work together and that you can access as a single system

The individual computers in a cluster act together to provide:

• Distribution of processing load

• Automatic recovery from failure of one or more components in the cluster

5

Discussion: Why Use Clusters? Can you realize a return on investment when you deploy an expensive cluster?What issues might you encounter when network services are not available?Which services would you make highly available?

Clustering TerminologyTerm DescriptionNode A server that participates in a cluster Resource A device or service hosted on a cluster and

accessed directly or indirectly by the application or end user

Failover clustering

An highly available clustering type with resources owned by a single server at a time

Load balancing A clustering type that distributes processing across a number of nodes

Fault tolerant A key component of clusters that withstands problems in hardware or software while continuing to operate

Planned downtime

The amount of time an application is unavailable due to an update or other maintenance

Unplanned downtime

The critical amount of time an application is unavailable due to a component failure

7

What is High Availability?The world is now a 24/7 global marketplaceSystems must be online or customers are lostGoal of high availability (HA) is to keep systems, applications, services, email, databases, files & printers readily availableEvery business now has high availability needs

Uptime Percentage

Maximum downtime per year

99.999 5 minutes99.99 52 minutes99.9 8.7 hours99 3.7 days

8

Why is HA Important?Server downtime is unavoidableKeep your business running and competitive

Servers may go offline due toMaintenanceUpgrade

Software or HardwareUpdate

Hot Fix, Security PatchAccidentPower OutageDisasters

Start planning now!

9

Introduction to Failover ClusteringIntroduction to Clustering

What is clusterWhy use clusterCluster common terminologyWhat is High AvailabilityHardware / Software RequirementsCluster Validations / Cluster Creations.

Introduction to WNLBOverview of Network Load Balancing Clusters Configuring a Network Load Balancing Cluster Maintaining a Network Load Balancing Cluster

10

Failover Clustering2+ machines (nodes)Redundancy everywhere

Storage, NICs, HBAs, MPIO, etc.“Shared” storage accessible by all nodes1 node will host a HA applicationApplication writes data to storage

11

Failover ClusteringNodes monitor health of other nodesIf that node fails, health monitoring will cause a “failover” of the resourceAnother node starts the application and reads the last saved information from the storageClients experience a slight interruption in service

12

What is a Failover Cluster?

Public

Shared Storage

HA Roles

13

Terminology Changes

Wolfpack Microsoft Cluster Service

(MSCS)

Server Clustering Failover Clustering

(WSFC)

Beta 1997 2000 2003 2008 R2

14

Software RequirementsClustering comes as an in-box feature on:

Windows Server 2008 & R2 DatacenterWindows Server 2008 & R2 EnterpriseWindows Server 2008 & R2 for IA-64Microsoft Hyper-V Server 2008 R2Windows Unified Data Storage Server Enterprise

Architecture:x64: up to 16 nodesx86: up to 8 nodesIA64: up to 8 nodes

15

Building Your Cluster

2 or more computers (nodes)2 NICs + dedicated storage adapter

3rd NIC for iSCSIHBA

3 NetworksPublicPrivate (heartbeat)Storage / iSCSI

Shared StorageOS, Service or Application

HA Roles

16

Mix And Match Hardware You can use any hardware configuration if

It passes ValidateEach component has a “Certified for Windows Server 2008” logo

Servers, Storage, HBAs, MPIO, DSMs, etc…

It’s that simple!Connect your Windows Server 2008 logo’d hardwarePass every test in Validate

It is now supported!If you make a change, just re-run Validate

Details: http://go.microsoft.com/fwlink/?LinkID=119949

17

Clustering Storage

• SCSI-3 SPC-3 compliant SCSI Commands• Persistent Reservations (PRs)• Parallel-SCSI deprecated in 2008• Multipath IO (MPIO) recommended• Basic GPT and MBR disks supported

Fibre Channel iSCSI SAS

Windows Server 2008/R2 Supported Shared Bus Types:

18

Networking

Key clustering componentPublic network – clientsPrivate network – cluster communicationStorage network – nodes access “shared” storage

Multiple networks for added redundancyIntegrated with new TCP/IP StackFull IPv6 SupportDHCP Support for IPv4 and IPv6 ResourcesNodes can reside in different subnets

19

Introduction to Failover ClusteringIntroduction to Clustering

What is clusterWhy use clusterCluster common terminologyWhat is High AvailabilityHardware / Software RequirementsCluster Validations / Cluster Creations.

Introduction to WNLBOverview of Network Load Balancing Clusters Configuring a Network Load Balancing Cluster Maintaining a Network Load Balancing Cluster

20

Validating a ClusterFor Microsoft support, cluster must pass the built-in Validate a Cluster Configuration (Validate) test

Run during configuration and/or after deploymentBest practices analyzed if run on configured cluster

Series of end-to-end tests on all cluster componentsConfiguration info for support and documentationNetworking issuesTroubleshoot in-production clusters

More information http://go.microsoft.com/fwlink/?LinkID=119949

21

Validate Inventories

• Same OS version which supports clustering• Same Hotfix and Service Pack level

OS BinaryConsistency

• CPU architecture• Memory informationArchitecture

• Domain membership and role• Analysis of unsigned driversConfiguration

• PnP device inventory• HBAs and NICsDevices

22

Validate Verifies

• Inter-node communication• SCSI compatibility with Persistent Reservations

Infrastructure

• Multiple NIC’s per server• Shared disks accessible from all machines and

uniquely identifiableHardware

• Each NIC has different IP address on a dissimilar subnetSoftware

• Network and Disk I/O latencies• Failover simulationFunctionality

DemoCluster Validation

24

WS03 – Cluster Creation

25

WS08 & R2 – Cluster Creation

DemoCreating a Cluster

27

Windows Network Load Balancing (WNLB)

Overview of Network Load Balancing Clusters Configuring a Network Load Balancing Cluster Maintaining a Network Load Balancing Cluster

28

What Is Network Load Balancing (NLB)?

NLB: • Enables high availability and scalability for Internet server-based applications • Enables clients to access two or more servers using a single IP address • Requires all servers in the NLB cluster to be running the same server applications with the same configuration

29

How NLB Works

When using NLB: • All servers in an NLB cluster are accessible by using a single (virtual) IP address • Client requests are distributed across available servers in the NLB cluster based on a common algorithm • All servers monitor each other through heartbeat messages • If a server fails to send heartbeat messages for 5 seconds, the other servers automatically converge and redistribute the client connections across the available servers

30

How Network Load Balancing Clusters Work

LAN

(Eth

erne

t)

NLB host +IIS server

Clients

NLB Web server cluster

NLB host +IIS server

NLB host +IIS server

31

Using Failover Clusters with Network Load Balancing

Clients

NLBWeb servers

Failover clusterdatabase server

32

Microsoft® Windows Server® 2008 NLB Features

• Network connectivity with support for:

- IPv6- Network Driver Interface Specification (NDIS) 6.0

- Multiple dedicated IP addresses per node

• Support for up to 32 computers per cluster • Denial of service attack and timer starvation protection • Support for rolling upgrades from Windows Server 2003

• Network Load Balancing Manager

33

NLB Deployment Requirements

To deploy network load balancing: Configure one or more network cards on all servers in the cluster

ü

Configure DNS to resolve the shared IP address ü Install and configure the same applications on all hosts in the cluster

ü

Deploy servers with similar hardware or plan for variations in load balancing

ü

Install the Network Load Balancing feature ü

DemoCreating NLB Cluster

35

Considerations for Configuring NLB Clusters and Hosts

When configuring an NLB cluster, you should consider whether: • Each server will host the same number of client connections • The client will connect to the same server for all connections • The cluster will operate the same way for all applications • Your network switches support unicast or multicast operations

Comparing Unicast and Multicast NLB Cluster Operation Modes

The cluster operation mode is used to ensure that all cluster hosts receive all incoming client requests sent to the virtual IP address assigned to the cluster

Method Description

Unicast

• All cluster hosts share an identical unicast media access control (MAC) address

• The outgoing MAC address is modified, based on the cluster host’s priority setting, to prevent switches from discovering that all cluster hosts have the same MAC address

Multicast

• Each cluster host retains the original adapter MAC address• The adapter is assigned a multicast MAC address, which is shared by

all cluster hosts

Options for Configuring Host Parameters

Parameter Description

Priority • Specifies a unique ID for each host• Used to assign one server to handle all network

traffic not covered by a port rule

IP address • Must be unique in the cluster

Subnet mask • Must be the same for all hosts

Initial host state • Specifies whether NLB will start and whether the host will immediately join the cluster

38

What Are Port Rules?Port rules are policies that define how to direct client requests to cluster hostsA default port rule applies to all protocols

Port rules define: • The cluster IP address where the port rule is applied • Port range and protocol (TCP, UDP, or both) • Filtering for multiple hosts with affinity or single host

• Load weight

• The option to disable a port range

39

What Is Affinity?Affinity is a configuration option that you use to distribute incoming client connections

Affinity options include:• None: Allows multiple connections from the same client IP address to be handled by different cluster hosts • Single: Specifies that NLB direct multiple requests from the same client IP address to the same cluster host • Network: Specifies that NLB directs multiple requests from the same IPv4 Class C or IPv6 network address range to the same cluster host

40

ResourcesCluster Team Blog: http://blogs.msdn.com/clustering/ Cluster Information Portal: http://www.microsoft.com/windowsserver2008/en/us/clustering-home.aspx Clustering Technical Resources: http://www.microsoft.com/windowsserver2008/en/us/clustering-resources.aspx Clustering Forum (2008): http://forums.technet.microsoft.com/en-US/winserverClustering/threads/Clustering Forum (2008 R2): http://social.technet.microsoft.com/Forums/en-US/windowsserver2008r2highavailability/threads/ Clustering Newsgroup: https://www.microsoft.com/technet/community/newsgroups/dgbrowser/en-us/default.mspx?dg=microsoft.public.windows.server.clusteringFailover Clustering Deployment Guide: http://technet.microsoft.com/en-us/library/dd197477.aspx TechNet: Configure a Service or Application for High Availability: http://technet.microsoft.com/en-us/library/cc732478.aspx TechNet: Installing a Failover Cluster: http://technet.microsoft.com/en-us/library/cc772178.aspx TechNet: Creating a Failover Cluster: http://technet.microsoft.com/en-us/library/cc755009.aspxWebcast: TechNet Webcast: Build High-Availability Infrastructures with Windows Server 2008 Failover Clustering

41

42

Q & A

STEP Event Survey

http://tinyurl.com/STEP140

More details join to http://www.bitpro.in