The health state of the cluster and the nodes deployed within the cluster is evaluated based on the cluster health policies. Custom cluster health policies can be set in the cluster manifest file under the FabricSettings section. The following is an example:

<FabricSettings> 
  <Section Name="HealthManager/ClusterHealthPolicy"> 
    <Parameter Name="ConsiderWarningAsError" Value="False" /> 
    <Parameter Name="MaxPercentUnhealthyApplications" Value="20" /> 
    <Parameter Name="MaxPercentUnhealthyNodes" Value="20" /> 
    <Parameter Name="ApplicationTypeMaxPercentUnhealthyApplications-
                     ControlApplicationType" Value="0" /> 
  </Section> 
</FabricSettings> 

Let's now understand what each of these parameters listed in the preceding policy is used for:

• ConsiderWarningAsError: This parameter specifies weather or not to treat warning reported by the health monitors as errors. This parameter is set to false by default.
• MaxPercentUnhealthyApplications: This parameter decides the threshold for marking the cluster entity as unhealthy based on the number of applications which can afford to be unhealthy at a given point of time. The default value for this entity is 0.
• MaxPercentUnhealthyNodes: This is the threshold for marking the cluster as unhealthy based on the number of unhealthy nodes at that given point of time. This parameter helps us accommodate for nodes which are undergoing maintenance.
• ApplicationTypeHealthPolicyMap: This parameter helps qualify special applications with specific rules. In the preceding example the parameter is used to specify a tolerance threshold for applications of the type ControlApplicationType to 0. This feature helps specify significance of some applications over other in terms of their expected availability.