4.1.0 HP PolyServe Software for Microsoft SQL Server administration guide (T5392-96074, October 2010)

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, HP recommends 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, HP PolyServe Software 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, HP PolyServe Software can attempt
to restart the service. (When you create the monitor, you can specify the number of
times that HP PolyServe Software 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.
Introduction12