15.7 multi-subnet failover clusters

15.7 multi-subnet failover clusters

15.7.1 Multi-subnet

SQL Server 2012 introduces a stretch cluster set (Stretch Clusters) that enables clustering across geographically dispersed sites. Each failover cluster node is connected to a different subnet (or subnet group) and is therefore referred to as a multi-subnet failover cluster.

In a multi-subnet cluster, there is no shared storage accessible to all nodes, so data storage needs to be replicated between multiple sub-networks. This deployment mode, which makes the database have multiple replicas, provides remote disaster recovery capabilities.

650) this.width=650; "src=" Http://s3.51cto.com/wyfs02/M02/6E/4A/wKiom1V4CkzAE7tPAAD7XJlm9qA296.jpg "title=" topology. png "alt=" Wkiom1v4ckzae7tpaad7xjlm9qa296.jpg "/>

When one of the IP addresses on one of the primary nodes fails, failover does not automatically occur until all the IP addresses that are valid on that node fail. When a failover occurs, SQL Server enters the online state if it can bind to at least one IP address that is valid on the current node.

15.7.2 IP Address Resource

In Windows Server 2003, the dependencies between the resources of the failover cluster are both and. In SQL Server R2 and previous versions, SQL Server will traverse all IP addresses in the failover cluster at startup and attempt to bind them all. If the binding fails, the startup of SQL Server also fails. Therefore, SQL Server can only enable multi-site failover clusters with stretched VLANs.

Windows Server 2008 allows a dependency between the resources of a failover cluster to be set to and OR OR. When you install a cluster, SQL Server 2012 can intelligently detect multi-subnet environments and automatically set IP address resource dependencies to OR. In addition, SQL Server 2012 will no longer need to bind to all IP addresses that are not online when it is started.

650) this.width=650; "src=" http://s3.51cto.com/wyfs02/M01/6E/44/wKioL1V39F-jIWL6AAQquHUEHYI885.jpg "title=" IP address Offline. png "alt=" wkiol1v39f-jiwl6aaqquhuehyi885.jpg "/>

15.7.3 Client Connections

The client must use the new parameter in its connection string: Multisubnetfailover=true.

By default, the client application retrieves all registered IP addresses from the DNS server, attempts to connect to them sequentially or in parallel, and connects to the first responding server.

For older client drivers or third-party data providers, because the Multisubnetfailover parameter cannot be used in the connection string, it is recommended that the time-out period of 15 seconds be increased for each additional IP address in order to prolong the client connection. For example, the default connection time-out period is 15 seconds and the failover cluster has 3 sites with IP addresses, so the new ConnectionTimeout value is recommended for: 15 + (3-1) * 21 = 57 seconds.

This article from "SQLServer2014 series" blog, declined reprint!

15.7 multi-subnet failover clusters