3.4.3 MxDB for SQL Server Installation and Administration Guide

Chapter 1: Introduction 4

This configuration supports only one failover at a time. If the Virtual SQL

Server on node 1 (“virtual1”) fails over and node 2 then goes down, the

Virtual SQL Server on node 2 (“virtual2”) will not be able to fail over. This

occurs because “instance1” on node 4 is currently providing access to the

databases for the “instance1” associated with the Virtual SQL Server

“virtual1.”

A similar situation occurs when a Virtual SQL Server is moved to

maintenance mode (for example, to apply a service pack). In maintenance

mode, all SQL instances associated with the Virtual SQL Server are shut

down on the primary and backup nodes. In the above example, if the

Virtual SQL Server on node 1 (“virtual1”) is moved to maintenance mode,

“instance1” on node 1 and also its backup “instance1” on node 4 will be

shut down. There is now no backup available for “instance1” on node 2

and node 3. To minimize the time without a backup, we recommend that

you perform the maintenance on the backup instance first and then

reenable it. The backup instance will then be available for the other

primary instances.

SQL Monitors and Failover

When you add a SQL instance to a Virtual SQL Server, MxDB for SQL

Server creates a service monitor for that instance. The monitor is active on

the node currently hosting the Virtual SQL Server and periodically

determines whether the SQL service is up.

If the monitor detects that the SQL service is down, MxDB for SQL Server

can attempt to restart the service. (When you create the monitor, you can

specify the number of times that MxDB for SQL Server should try to

restart the service.) If the attempts are not successful, by default the

Virtual SQL Server will fail over to a backup node.

You can customize the failover behavior for each SQL service monitor.

For example, you can use the Policy configuration option to specify

whether the Virtual SQL Server should fail back to the original node or

remain on the backup node after the problem causing the failover has

been corrected.