IBM Tivoli Netcool/OMNIbus Version 8.1

Monitoring predictive events in the event list

A predictive event represents a prediction of an impending threshold violation. These events differ from other events that are forwarded to Tivoli Netcool/OMNIbus because they represent an occurrence, fault, or situation that has not yet occurred, but which has a significant probability of occurring in the near future.

About this task

Predictive events are generated within IBM Tivoli Monitoring based on linear trending of historical data. You can monitor predictive events from the Tivoli Netcool/OMNIbus event list if your system is configured to support this feature. If configured also, a Predictions monitor box will be available in the Event List monitor box window, and a Predictions filter and Predictions view will be available in your event lists. The Predictions view contains the following default columns, which are displayed as follows, from left to right: Node, TrendDirection, Summary, FirstOccurrence, LastOccurrence, Count, PredictionTime.

The sort priority and sort order of the columns is as follows:

Severity in descending order
LastOccurrence in ascending order
PredictionTime in ascending order

When your event list shows both actual and predictive events, you can use the Predictions filter and Predictions view to filter and sort predictive events. You can also identify predictive events as follows:

Check the text in the Summary column. The summary text is typically in one of the following formats (the severity of critical predictive events is always 5, the severity of warning predictive events is always 3):
- For linear trending events:
  Predicted event rate on Node will hit threshold within PredictionTime days.
- For baselining events:
  EventType event rate baseline situation has fired for node Node, the event count is EventCount for node situation_name.
- For predictive events:
  Critical threshold for resource on Node will be breached in PredictionTime day(s). (Situation: situation_name).
  
  Warning threshold for resource on Node will be breached in PredictionTime day(s). (Situation: situation_name).
If you have a Class column in your current view, check the text in this column. By default, the entry is Predictive Events, although this value can be changed by your system administrator.

From the event list, you can perform the following operations on predictive events:

Procedure

Determine the confidence level of the predictions. To display predictive events that meet a certain prediction confidence level:
1. Right-click any event in the event list, and then click Show Prediction Confidence from the pop-up menu.
2. From the Internal Command Parameters window, specify a percentage value between 0 and 100 only, in the Minimal Confidence percentage field. The prediction confidence indicates how accurate the approximated trended value is, where 0 is no confidence and 100 is a perfect approximation. The default is 50.
3. Click OK. A separate event list opens displaying predictive events with the specified, or higher, confidence level. This event list shows all the fields in the Default view. You can additionally double-click any event to view its complete set of details in the Event Information window.
List all predictive events and actual events (if available) that originated from the host for which a prediction was made. You can use the results displayed to help you identify a possible cause for the prediction trend towards a warning or critical threshold.
To display the events, right-click the predictive event that you want to investigate, and then click Show Predictive Events Threshold from the pop-up menu. A separate event list opens showing all events that are on the same host as the selected predictive event, and which occurred from the time the prediction was made, to the time the prediction is calculated to occur. (The Default view is applied to this event list.)

You can double-click each event to view its complete set of details in the Event Information window.

Library | Support |