You can use Root Cause Analysis (RCA) to filter a set of events to determine the cause of a higher level system, service, or application problem. RCA can help you to eliminate apparent performance problems that may otherwise appear to be root causes but which are only side effects or symptoms of the actual root cause of the problem, allowing you to more quickly identify problem areas. You can view the RCA results on the home page or topology page of any service that is currently down. The topology page allows you to see a graphical representation of the service, system and component dependencies with the target(s) highlighted that RCA has implicated as causing the service failure.
The metric specified will be tested for this component during Root Cause Analysis upon service failure. Specify the object (if applicable) and the threshold to use for the test. You may use the critical alert threshold for this metric, or you may specify an alternate threshold to be used for Root Cause Analysis. The threshold you specify, when exceeded, should indicate a fault condition for this target as a key component of the specified service.
You use these metrics in conjunction with both the availability of the the target itself and a selection of the metric tests you define against that target. Once the metrics are defined, the status of the target during Root Cause Analysis depends not only on the status of the target, but also on the status of the metric tests performed against that target. When RCA runs, the target and the metric tests are treated such that if any of them indicate a system failure, then the service that depends on that test also exhibits a system failure.
If you specify a fault threshold for a metric test included in the dependency topology, RCA is still driven by the presence of an alert on that metric. However, the presence of the alert alone does not indicate that the test has failed. RCA retrieves the current value for the metric and compares it to the fault threshold provided. If the metric exceeds the threshold then the test is considered to be at fault. If an alert occurs for a metric in this situation and the value for the metric does not exceed the fault threshold, then the test is not considered to be at fault, and notification will proceed for the alert.
Note: You can define a component test for any target included as a direct dependent of a service.
Related Topics
About Root Cause Analysis Processing
Getting the Most From Root Cause Analysis
Enterprise Manager Information Roadmap
Copyright © 1996, 2005, Oracle. All rights reserved.