This page provides details on insights in Stackdriver Incident Response and Management (IRM).
IRM insights help you accelerate your incident investigation process and ultimately reduce your incident mitigation and resolution time.
Insights point you to the potential triggers and contributing factors for an alert, steering your investigation in an appropriate direction and enabling you to address the root-cause issues.
Where available for a given alert, insights are shown in the Alert Details view in an Insights panel under the alerting metric chart. For example, your alert might contain metric correlation insights:
No insights to display indicates that IRM currently has no insights for this alert. In some cases, insights might be discovered later and will be displayed automatically.
Types of insights
The following section lists the types of insights currently available in IRM. These insights, if available for an alert, are generated automatically; you don't need to configure them.
Metric correlation insights identify a metric that is strongly correlated (positively or negatively) with the time series that triggered the alert. The correlated metric might indicate a cause of the problem, or possibly a secondary effect that didn't trigger an alert.
A predetermined set of metrics is used for the correlation.
Anomalous behavior insights identify any Google Cloud Platform services that exhibited anomalous behavior near the time the alerting policy was triggered. These insights indicate that the problem might have been caused by an issue in the GCP service.
For instance, an anomalous behavior insight might include the following type of statement:
spanner.googleapis.com service at us-east1 experienced recent
anomalous behavior in the
for the request_latency_50 health metric."
Alerting policy modified
Alerting policy modified insights report whether the alerting policy that triggered the alert was recently modified. These insights indicate that the policy might have been misconfigured, thereby triggering a false positive alert, or that the condition was pre-existing.
Use insights to investigate an alert; for details, go to Investigating an alert.