Downloaded 133 times
Uploaded byJoseph D'Antoni
Building your first sql server cluster
This document outlines the steps for building a SQL Server cluster for high availability, including planning considerations, required hardware, installing Windows clustering features, configuring storage, installing and configuring SQL Server across nodes, and testing the cluster configuration. Key aspects that are discussed include defining recovery time and point objectives, installing SQL Server using the "Create New Failover Cluster" option, installing SQL on each node to enable failover, and performing backups and restores from cluster-owned drives. Testing the applications on the clustered environment is also emphasized.
Building your first sql server cluster
- 1. Building your firstSQL Server ClusterJoe D’AntoniSQL Saturday #69Philadelphia05 March 2011
- 2. DBA with 10+years of experience
- 3.
- 4. Vice President ofthe Philadelphia SQL Server User Group
- 5.
- 6.
- 7. High availability—why doyou need it?Architectures—HA and DRPlanning—who needs to be in the room?Definitions—It helps to have a glossaryTechnical—what do you need to do?Support—How is supporting this environment different from supporting a standaloneAgenda
- 8. Why do youneed a HA environment?Remember HA is not DR.High Availability
- 9. SQL Server ClusteringSQLServer MirroringPeer to Peer ReplicationSQL Server Log Shipping*High Availability Options in SQL
- 10. Backup and RecoveryMirroringLogShippingReplicationSAN Replication*Virtualization*DR Options in SQL Server 2008
- 11. Hardware ProtectionO/S ProtectionHighavailabilityClustering - Why
- 12. Depends on sizeof your IT organizationWindows System AdminsDBAsStorage AdminsNetwork AdminsIdeally—application leadsClustering—Who
- 13. RTO—Recovery Time Objective.How long can your systems be down before impacting the businessRPO—Recovery Point Objective. How much data can you lose before affecting the business.Cluster—The Windows cluster (consisting of 1 or more nodes) that your SQL Server instance runs on top ofResource Group—services and disks that are associated with your clustered service (in this case SQL) Multi-instance/single-instance—Official MS terms for active-active and active-passive clustersSome Definitions
- 14. SAN—storage area network,a box of many disks which can be presented to multiple serversLUN—Officially Logical Unit Number, but practically a disk volume presented from SAN to a server(s)Mount Points—Naming method for Windows disk devices, attaching many devices to single drive letterNode—The physical (or virtual) Windows machine supporting your clusterMore Definitions
- 15. Failover Cluster Manager—Theutility in Windows that provides for cluster management and verification.Quorum disk—Disk that verifies all of the nodes in the cluster can talk to each otherLooksalive/Isalive—Processes which verify if cluster services are still running and initiates failoverMore Definitions
- 16. Two ServersSANNetworkIP AddressesWhatyou need to build a cluster?
- 17.
- 18.
- 19. Almost all SANswill doYou may need to update your SAN firmwareWindows 2008R2 Requires iSCSI-3 persistent reservationsWork closely with your storage team, to ensure proper disks (mainly for performance)LUNs must be presented to all cluster nodesSAN
- 20. Add Failover Clusterfeature to your server nodesWindows FirewallAntivirusValidate your clusterName your cluster, and reserve its IP addressStart using mount points to label your disks!!!!In a multi-instance cluster, it’s really easy to run out of drive lettersWindows
- 21.
- 22. Mount Points --Windows
- 23. Storage Network +Server/Windows + Switch
- 24.
- 25. Many, many IPaddresses and aliasesHeartbeat NetworkDTC ClusternameInstance ServiceOf course, the base nodeVerify ports are not blockedNetworkI would put this slide after the windows one…I would build from the physical (SAN, Servers/Windows, switch), to the “less physical” (network)
- 26. Storage Network +Server/Windows + Switch + IPs
- 27.
- 28. Cluster DTC Service(requirement)Slipstream SP1 installationInstall SQL using “Create New Failover Cluster Option”Install first on the node that is the current owner of the disks you would like to useMust install SQL on each nodeSQL Server 2008 Install
- 29. This should bepretty easy!!Select the add node option—then select the instance you just createdYou will then have to specify passwords for the servicesInstallation should be completeDo a failover for test purposesSecond Node Installation
- 30. Memory pressure inmulti-instance SQLBackups and Restores must be performed from drives owned by the cluster serviceThis is more of a hardware consideration—but HBA trafficRemember—this isn’t a shared everything environmentConsiderations for SQL Server
- 31. Performance is generallythe same as in a standalone instanceTest your applicationsSSIS—use MSDB storage and rely on server for securityApplication Considerations
- 32. Windows UpdatesFirmware UpgradesSQLServer CPs and SPsLeveraging your Cluster
- 33. Understand your HAand DR prioritiesKnow who needs to be involved in planning processBuild your clusterTest your applications!Summary
- 34. Great book—Pro SQLServer 2008 Failover Clustering by Allan HirtPaul Randal—White Paper on HA Solutions in SQL 2008Slipstreaming SQL Server InstallBuilding a cluster from MSBibliography
- 35.
- 36.