Database Reference
In-Depth Information
Figure 31-1.
Two-node WSFC with a single SQL Server Cluster instance
One of the SQL Server instances is active and handles all user' requests. Another node provides hot standby.
When something happens to the active node, the SQL Server cluster fails over to the second node, formerly the
passive
node, and starts from there. This process, in a nutshell, is a SQL Server instance restart. The new active node
performs a crash recovery of all of the databases in the instance, preventing clients from connecting to the databases
until this process is complete.
The duration of crash recovery and failover greatly depends on the amount of data modified by active
transactions at the time of the failover. With short OLTP transactions, failover could be in the under a minute range.
However, it is possible that failover can take much longer in the case of active transactions that modified a large
amount of data and needed to be rolled back by a crash recovery process.
A SQL Server Failover Cluster fails over the entire SQL Server instance. It is impossible to have some databases
running on a SQL Server instance installed on one node of the cluster and other databases running on another SQL
Server instance installed on a different node.
Failover Clustering requires that all databases be placed into shared storage. Therefore, storage becomes the
single point of failure.
■
always use highly redundant storage with Failover Clustering. Moreover, consider combining a sQl server
Failover Cluster with other high availability technologies that allow you to store another copy of the databases on a different
storage device. it increases the availability of the system and minimizes possible data loss in case of a storage failure.
Important
