747
Chapter 29: Clustering
29
of SQL Server 2012 instances that are used for reporting purposes. You would replicate your
data out from a single SQL Server 2012 instance to multiple servers that reside behind some
kind of hardware or software load-balanced solution. This would provide you with a great
deal of computing power to handle users’ queries while providing you with a level of report-
ing redundancy (where individual machines behind the load balancer could go down and
others would still be available to handle traffi c).What Does Online Mean?
SQL Server 2012 is considered as being online when the service is up and running and you can
connect. For the SQL Server services to come up and remain online, there are several resources
also required to be online. A failure in any one of these dependencies would cause SQL Server
to go offl ine and the cluster to failover to another node. These critical resources include:■ (^) Network name
■ IP address
■ (^) Shared disk storage
If for any reason any of the required resources fail and cannot be brought up, SQL Server
shows as offl ine in the Failover Cluster Manager, and you cannot connect:
How Does Clustering Work?
Clustering functions by performing constant checks to ensure that nodes, drives, IP address,
network names, and many more items are up and available. This constant check of both the
Windows and SQL Server resources serves to ensure that SQL Server remains available.
By default, Windows 2008 Failover Clustering performs a basic health check on each of
these resources every 5 seconds and a more thorough check every 60 seconds. Both of these
values are confi gurable from within the Failover Cluster Manager.
When the health check runs again SQL Server, resources also query the spserver
diagnostics stored procedure. By default, the health check expects results within 60
seconds. (Although this can be confi gured and set as low as 15 seconds.) If no results are
returned, by default, the cluster initiates a failover of the SQL Server instance to another node
in the cluster. This action is also confi gured by adjusting the value of FailureConditionLevel for
the SQL Server resource. There are six levels from which you can choose:
■ 0 : No automatic failover or restart
■ (^1) : Failover or restart on server down
■ Triggered if the SQL Server service is down
■ (^2) : Failover or restart on server unresponsive
■ Triggered on level 1 error OR
■ (^) Timeout from sp_server_diagnostics
c29.indd 747c29.indd 747 7/31/2012 9:51:58 AM7/31/2012 9:51:58 AM
http://www.it-ebooks.info