Designing and Managing Behavior Models - Behavior Models and Their Components - Constructing Behavior Models - How the Pieces Fit Together -
Previous: Constructing Behavior Models     Next: An Example of a Behavior Model

How the Pieces Fit Together

Let's first review how you define which managed nodes a behavior model will monitor and manage. As the figure below shows, each node belongs to a property group, and that property group contains properties.

Nodes, Property Groups, and Properties

Any set of nodes that share a unique property can be managed as a set of devices. (The nodes need not be members of the same property group.) In the figure above, the tcp property might be that unique property.

For a node to be pollable, the principal requirements are that:

• The poll's property must be in the node's property group.
• The base object around which the poll's poll condition is built must be a property in the node's property group.
• The poll's trigger must correspond to a pending alarm transition, and the alarm's property must be in the node's property group.

The figure below shows the definition of a poll that has been designed to work with the node shown in the previous figure.

Relationship Between Node and Poll

As you can see, the node's property group, Mib-II, contains a property tcp that matches the poll's property and the base object used in the poll's poll condition. Once this poll is enabled, the poll TcpMedRetrans will poll the node, unless there is no alarm that the poll can affect or the node is suppressed. (If the node is suppressed, no polling will occur because the poll is marked suppressible.)

---

Note    Since trap masks do not have properties, this type of matching is not necessary for masks.

---

If TcpMedRetrans polls the node, receives a response to its query, and that response satisfies the poll condition, the poll will fire a trigger. If an alarm has been defined whose first transition is tcpRetransMed (the poll's trigger) and that alarm is enabled has the property tcp, a new instance of that alarm will be instantiated to monitor the node. Because the alarm is instantiated using the trigger's Node and Subobject, the key attributes of the trigger and alarm will match, and the first transition will be effected.

Once an alarm instance has been instantiated and has gone through one transition, the transitions that can be effected from its current state determine which triggers affect the alarm. For example consider the following alarm, TcpRetransMon.

An Alarm: TcpRetransMon

When this alarm is first instantiated and the tcpRetransMed transition is made, the alarm transitions to the tcpMedRtrans state, so two transitions are pending: tcpRetransNorm and tcpRetransHigh. If NerveCenter sees a trigger with one of those names, and the trigger's Node and Subobject match those of the transition, the transition occurs.