SQL Server 2008 : Failover clustering - Installing a clustered SQL Server instance

The process of installing a clustered SQL Server instance has changed since SQL Server 2005. There are now two installation options: Integrated and Advanced.

1. Integrated vs. advanced installation

An integrated installation creates a single-node failover cluster, from which additional nodes (nodes on which the instance can fail over to) are added via a separate installation. As shown in figure 2, the initial and subsequent node installations are started by choosing the New SQL Server Failover Cluster Installation and Add Node to a SQL Server Failover Cluster options on the Installation tab of the SQL Server Installation Center.

In contrast to the one-node-at-a-time approach of the integrated installation, the advanced installation prepares multiple-cluster nodes in one step, before completing the installation on the node chosen as the initial active node for the instance. As shown in figure 3, you specify this installation type by selecting the Advanced Cluster Preparation and Advanced Cluster Completion options on the Advanced tab of the SQL Server Installation Center.

Figure 1. The Possible Owners setting gives you control over the cluster failover process.
 

Figure 2. Choose the New SQL Server Failover Cluster Installation option to create a single-node failover cluster before adding additional nodes with the Add Node to a SQL Server Failover Cluster option.
 

Figure 3. Choose the Advanced Cluster Preparation and Advanced Cluster Completion options to streamline the process of installing a clustered SQL Server instance on multiple-cluster nodes.
 

Clustered installations share some of the same installation screens and steps, so rather than repeat them, let's walk through the steps unique to a clustered installation using the integrated method.

2. Integrated installation steps

As with a nonclustered installation, you begin a failover clustering installation by running setup.exe from the installation DVD. Next, you go through a series of steps to install setup support files and check various setup rules. As shown in figure 4, the setup checks for a clustered installation are more detailed than for a nonclustered installation.

Installation continues with the usual prompts for a product key, acknowledgment of license terms, and feature selection, before arriving at the instance configuration step, as shown in figure 5.

The one difference between this step and the equivalent step in a nonclustered installation is the SQL Server Network Name field. The name you enter is used to identify an instance on the network. In our example, we'll use BNE-SQL-PR-02 as our network name, and together with the instance name (Marketing), we'll access this instance as BNE-SQL-PR-02\Marketing without ever needing to know which of the two cluster nodes the instance is running on.

Figure 4. The Setup Support Rules for a clustered installation include cluster-specific checks such as the existence of a clustered Microsoft Distributed Transaction Coordinator (MSDTC) service.
 

Installation continues through the disk space requirements check before prompting for a cluster resource group, as shown in figure 6. The resource group name is used as a container for holding the resources (disks, IP addresses, and services) for the installed instance.

Figure 5. A clustered SQL Server installation is identified on the network with a unique network name you specify during installation.
 

Figure 6. The cluster resource group name is used to both identify and group the instance's resources.
 

In the next step, you'll identify available cluster disk resources that can be chosen for inclusion in the instance's resource group. As shown in figure 7, the quorum disk and cluster disks that have been previously assigned to another clustered instance are unavailable for selection.

As shown in figure 8, the next step lets you specify either a static or DHCP-based IP address for the SQL Server instance.

Figure 7. On this screen, you identify available cluster disks for assignment to the instance's resource group.
 

Figure 8. Unlike earlier versions, SQL Server 2008 permits DHCP-assigned IP addresses.
 

Figure 9. The Cluster Security Policy tab lets you select between the default service SIDs and custom domain membership.
 

The only remaining cluster-specific installation step is configuring the Cluster Security Policy. In previous versions of SQL Server, the service accounts had to be added to newly created domain groups prior to installation. Permissions for the service accounts were then managed at the domain group level. This requirement was often misunderstood, and introduced complexities when the domain groups needed to be changed. In response to this, SQL Server 2008 introduced an alternative method that uses service security identifiers (SIDs).

As you can see in figure 9, using SIDs is the recommended configuration, although support for the old domain group method remains.

At the end of the installation, the clustered instance will be created and available, but can't fail over to other cluster nodes until you run the Add Node installation on the appropriate nodes. This installation option is used to enable additional cluster nodes to participate as failover nodes for an existing SQL Server failover clustering instance. Figure 10 shows one of the screens from this installation option, and in this case, we've chosen to allow the SALES2 server to host the MARKETING instance.

Figure 10. In this example, the Sales2 server is installed as a failover participant for the Sales instance installed on the Sales1 server.
 

Figure 11. Use the Cluster Management tool to view and manage a clustered SQL Server instance's resources and state.
 

When installation is complete, you can manage the clustered instance using the Failover Cluster Management tool in the Administrative Tools folder, or by running Cluadmin.msc from the Start menu. In the example in figure 11, you can manually move, or fail over, a clustered instance to another cluster node by right-clicking the resource group and selecting the "Move this service or application to another node" option.