On this page:
Related topics:
Anomaly Detection must be enabled, but requires no configuration except when you want to limit anomaly alerting. Enabling Anomaly Detection also enables Automated Root Cause Analysis.
Anomaly Detection and Automated Root Cause Analysis are available to SaaS customers only.
Enable Anomaly Detection
Enable Anomaly Detection separately for each application.
- In Alert & Respond > Anomaly Detection, choose the desired application from the drop-down menu, and Toggle Anomaly Detection to ON
Let Anomaly Detection Learn your Application
After you enable Anomaly Detection, it takes 48 hours for Anomaly Detection and Automated Root Cause Analysis to become available. During that time, the Machine Learning models train on the Business Transactions in your application.
To view Business Transaction training status:
- In Alert & Respond > Anomaly Detection, view the Model Training tab
The table below explains the five possible training statuses of a Business Transaction.
| Status | Meaning |
|---|
| In Training | Model training is in progress for the Business Transaction |
| Ready | Model training is complete, and the Business Transaction is healthy |
| Warning | Model training is complete, but the Business Transaction has experienced one or more Warning level anomalies during the training period |
| Critical | Model training is complete, but the Business Transaction has experienced one or more Critical level anomalies during the training period |
| Not Available | Model training is incomplete, and the Business Transaction is not visible to Anomaly Detection |
The models continue training as long as Anomaly Detection is enabled. If traffic to a Business Transaction is interrupted for long enough to prevent training that day, Anomaly Detection continues to function using the models from the previous 7 days.
No Machine Learning models are trained for Business Transactions which have very low CPM, because the sample size would be so small that the resulting model would be unreliable.
Monitor Anomalies
View the Applications > Business Transactions screen, and for any Business Transaction of interest, click the Warning or Critical icon in the Health column:
A list of Health Rule Violations and Anomalies for that Business Transaction appears:
Since there are multiple ways to view lists of anomalies, monitoring anomalies can be a natural extension of the way you work with AppDynamics.
Choose any of the following options to open a detailed view that includes the results of Automated Root Cause Analysis.
- In Alert & Respond > Anomaly Detection, view the Anomalies tab
If you monitor applications for an application ops team
- In Applications > Events, filter Event Types to include Anomalies
- In Applications > Troubleshoot > Violations & Anomalies, filter Event Types to include Anomalies
You can configure policies to be triggered by anomalies, similarly to the way you configure policies to be triggered by Health Rules.
By default, Anomaly Detection alerts you about all Anomalies for all Business Transactions in your application.
You can configure Anomaly Detection to only surface Anomalies for the combination of Business Transactions and severity levels that you specify. Do this if you prefer to see fewer, more narrowly focused alerts.
Automated Root Cause Analysis does not need to be enabled or configured.
To configure Anomaly Detection:
- Click Configure Anomaly Detection to open the configuration dialog
- If desired, limit the Business Transactions on which you want Anomaly Detection to alert
- Default (all Business Transactions) is shown
- If desired, limit alerts to either Warning or Critical severity only
- Default (All Severities, meaning both Warning and Critical) is shown
