Incident Prediction

InsightFinder’s Incident Prediction feature helps you predict potential incidents before impacting your business. Using your historical data to inform an adaptive algorithm, this feature flags data, assigns probability scores, and then estimates when an incident may occur. This predictive analysis shows not only the past data points that lead to the forecast but also provides actionable preventative measures. Without InsightFinder, these incidents would go undetected and cause service outages

Similar to InsightFinder’s Incident Investigation, Incident Prediction utilizes automatically created Knowledge Base rules, which are generated by root cause events of detected incidents and then tuned by historical predictions.

Explore a predicted incident

The top of the Incident Prediction page lets you filter through predicted incidents according to Component, Type (Matched, Unmatched, or Fixed), timeframe, or specific Pattern ID.

Select a specific predicted incident to see an overview including:

  • Matched Incident – Whether or not the incident matched a detected incident.
  • Instance – The instance at risk of the incident.
  • Pattern –The pattern ID categorizing with the incident.
  • Notification – Indicates an automated alert was sent to the operator.
  • Knowledge Base – Indicates the rule from Knowledge Base that created the prediction.

On the right side of the page, you can dig even deeper into the predicted incident’s likely root cause chain.

  • Prediction time – Time when the incident was predicted.
  • Predicted occurrence time – Time when the incident was predicted to occur
  • Patterns – The chain of anomalous patterns that led to the incident.
  • Triggered Action – Shows the automated action taken to remedy and fix the predicted incident.

NOTE: You can configure automated actions for predicted incidents in the Project Settings.

Incident

The Incident tab allows you to:

  • Set a pattern name
  • Mark the incident as “ignored”
  • Send an email alert to specific recipients
  • Export the incident details to a third-party integration.

Additionally, you can:

  • If the anomaly is a log anomaly, then open the anomaly in the Log/Trace Analysis page. If it is a metric anomaly, then open the anomaly in the Metric Analysis page.
  • Open the anomaly’s Pattern Trend.
  • Open the log query.

Further down the Incident tab, review the Lead Time Summary and Corresponding Detected Incident.

Recommendations

The Recommendations tab displays a table with Suggested actions along with the corresponding Component name and Instance name.

Operator Notes

The Operator Notes tab allows you to create, edit, and remove notes related to the predicted incident.

Triggered Actions

The Triggered Actions tab displays automated actions taken by InsightFinder to avoid the predicted incident.

See Also

  • Knowledge Base
  • Configure Actions
  • Incident Investigation