Lab: instrument, alert, and close the feedback loop

Outcome

Instrument a safe training application or telemetry sample, create an availability or failure alert, inspect the evidence, and turn one finding into a tracked work item. This completes the delivery loop from deployment to learning.

Before you begin

Use a training Application Insights resource and action group. Use a test email/webhook endpoint that you are authorized to notify. Do not expose a production endpoint or customer data for this lab.

Do

1. Enable Application Insights for a training app or use a supported sample that emits requests, dependencies, exceptions, and a custom feature_used event.
2. Create an availability test or a safe failure-rate alert. Route it to a test action group.
3. Query the telemetry with KQL or inspect the Application Insights experience to identify request count, failed requests, a dependency, and the custom event.
4. Create a work item/issue that links the alert or query evidence to a specific follow-up decision.

Verify

Capture the alert rule, action-group delivery, query/result, and linked work item. Explain the distinction between an infrastructure metric, an application request trace, a synthetic availability result, and a feature-usage event.

Break and fix

1. Trigger a safe failure or temporarily point the availability test at a controlled invalid route. Confirm the alert fires.
2. Remove the action from the test action group, confirm the signal can still fire without notification, then restore it.
3. Change the KQL/query time range so it becomes misleading; explain why and restore the useful query.

Scenario practice

1. Configure: Select the signal, alert threshold, action group, and issue owner for a new release whose failure rate must be detected quickly without paging on one transient error.
2. Diagnose: An alert fired but nobody was notified. Which evidence distinguishes the telemetry condition, alert rule, action-group routing, and notification delivery?
3. Distinguish: Contrast a metric, log query, distributed trace, synthetic availability result, and feature-usage event.

Clean up

Delete the test alert, action group, availability test, and training resources if they are no longer needed. Close the test work item with the recorded learning.

Sources and re-audit

• Application Insights availability tests and Azure Monitor alerts overview, checked 2026-08-22.

Re-audit when Application Insights telemetry, availability testing, KQL behavior, Azure Monitor alerting, or action-group delivery changes.

Related

telemetry · application-insights · azure-monitor · devops-flow-metrics · All labs