Dell PowerVault 775N (Rackmount NAS Appliance) - Manual (Page 16)

Table 1-7. Example of an N+I Failover Configuration for an Eight-Node Cluster
Cluster Resource Group Primary Node AntiAffinityClassNamesValue
A Node 1 AString
B Node 2 AString
C Node 3 AString
D Node 4 AString
E Node 5 AString
F Node 6 AString
Configuring Group Affinity
On N + I (active/passive) failover clusters running Windows Storage Server 2003, Enterprise Edition, some resource groups
may conflict with other groups if they are running on the same node. For example, running more than one Microsoft Exchange
virtual server on the same node may generate application conflicts. Using Windows Storage Server 2003, Enterprise Edition,
you can assign a public property (or attribute) to a dependency between groups to ensure that they failover to similar or
separate nodes. This public property is called group affinity.
Group affinity uses the AntiAffinityClassNamesValues public property, which ensures that designated cluster resources are
running on separate nodes, if possible.
For example, in Table 1-7 , the AntiAffinityClassNamesValues string for cluster resource group A and group B are identical
(AString), which indicates that these groups are assigned to run on separate cluster nodes, if possible. If node 1 fails, cluster
resource group A will failover to the next backup node (node 7). If node 2 then fails, because their
AntiAffinityClassNamesValues string value (AString) identifies group A and group B as conflicting groups, group B will skip
node 7 and instead fail over to node 8.
To set the public property for the cluster groups shown in Table 1-7 :
1. Open a command prompt.
2. Type the following:
cluster group "A" /prop AntiAffinityClassNames="AString"
3. Repeat step 2 for the remaining cluster groups.
Use the " Cluster Data Sheets " to configure group affinity in your N + I cluster configuration.
Failover Pair
Failover pair is a failover policy in which each application is allowed to failover between two specific nodes in a multinode
cluster. The Possible Owners List in Cluster Administrator is used to determine which nodes will run the failover applications.
This solution is easy to plan and administer, and any applications that do not run well on the same server can easily be
moved into separate failover groups. However, because failover pair uses one idle or passive cluster node for each cluster
pair, this failover option can be more expensive.
Figure 1-4 shows an example of a failover pair configuration. Table 1-8 provides a failover configuration for the cluster shown
in Figure 1-4 .