Previous Topic: Select Service Members

Next Topic: State Change Exits

Implementing Services › Service Definition Panels › State Thresholds

State Thresholds

From the Service Filters panel, press F8 (Forward) to go to the State Thresholds panel. Use this panel to define how the actual states of the members affect the actual state of the service.

The actual state of a service can be one of the following:

You must assign a threshold to the first four states.

Thresholds are evaluated in the order shown. The service takes on the state of the first threshold equaled or exceeded, irrespective of whether other thresholds are Equaled or exceeded. For each actual state, you specify a percentage threshold value that, if equaled or exceeded, causes the service to take on that state (unless a state of higher severity has also satisfied its threshold requirement). This threshold is expressed as a combined weight of the members required to deliver the service.

Each member of the service has a weight associated with it. The weight expresses the level of impact the individual resource has on the threshold calculation for the actual state of the service.

If members are not ACTIVE, you can use their logical state to calculate the threshold for the actual state of the service. In this case, if a member has a logical state of OK, its weight is added to the combined weight for the ACTIVE state. If a member has a logical state of UNKNOWN or STARTING, their weight is added to the combined weight for the corresponding actual state. If a member has any other logical state, their weight is added to the combined weight for the FAILED actual state.

Note: If a service filter finds no members, the weight specified in the Weight column on the Service Filters panel is added to the combined weight for the UNKNOWN state.

Using the logical state rather than the actual state to calculate the threshold has advantages. For example, you can shut down a resource that is part of a service without affecting the service. The service sees a logical state of OK, even though the resource is INACTIVE, and treats it as though it is ACTIVE. Alternatively, when a resource fails and you set it to IGNORED, the service sees the resource as ACTIVE (OK), and the service continues unaffected.