Dell PowerEdge R720xd Dell High Availability and Disaster Recovery Solutions U - Page 12

Dell High Availability and Disaster Recovery Solutions Using Microsoft SQL Server 2012 AlwaysOn Availability Groups Figure 4. Scenario 2 : SQL Server FCI with AVG on shared storage The advantage of the setup described within Figure 4 is, with the help of the read-only routing list, we may redirect or offload all read-only requests to Node3 and thereby reduce the overhead on the primary replica. Clients can see the latest data without any delay on Node3 as it is configured as Synchronous-Commit replica with primary replica. This is achieved at the cost of compromising the High Availability at the DR site. On the other hand, in the scenario described in Figure 3, we do not have a secondary node to serve the read-only requests in the primary site. However, we do have High Availability at the DR site. So it is up to the customer to choose which scenario best suits their requirements. In the scenario illustrated in Figure 4, we have installed the SQL FCI on Node1 and Node2 in the primary site. The main purpose of this FCI instance is to provide High Availability to the databases with automatic failover capability, and this is available for access as long as any of the two physical nodes is up and running. Now the Availability Groups can be configured over the FCI (SQL FCI running on Node1 and 2) and the standalone instances of other two nodes (Node3 and Node4). Since this is a 4-node cluster, we can configure this cluster with the "Node and Disk Majority" quorum setting. When disk witness is online, this quorum setting can sustain a maximum of any two node failures. The advantage of using the "Node and Disk Majority" quorum setting is that even when the FCI instance (if Node1 and Node2 goes down) is not available, we can still access the AVG databases through AVG listener on Node3 without needing to failover to DR site. Node4 is still available for manual failover for disaster recovery purposes. NOTE: In this configuration, automatic failover is achieved with the FCI instance which might be running on Node1 or Node2 at any point in time. Since the FCI instance does not support automatic failover/failback when we combined it with Availability Groups, we may have to do a manual failover to Node3. Since Node3 is Synchronous-Commit replica, failover can happen without any data loss. 12