Availability test: definition

Availability tests answer the question telemetry from inside the app can’t: can users reach it at all? A standard test sends an HTTP request on a schedule from up to 16 locations, and Microsoft recommends at least five. Beyond the response, a standard test can check the TLS certificate’s validity and remaining lifetime, use any HTTP verb, and send custom headers and a request body. A resource can hold up to 100 tests.

Each run becomes a row in availabilityResults (AppAvailabilityResults in a workspace), with the test name, location, duration and whether it succeeded:

availabilityResults
| where timestamp > ago(1d)
| summarize runs = sum(itemCount), failed = sumif(itemCount, success in~ ("false", "0")) by name, location
| order by failed desc

Standard tests are billed per run, $0.0005 in East US. A test every 5 minutes from 5 locations runs 1,440 times a day, about 43,200 times a month, which comes to $21.60. The older URL ping tests cost nothing; Microsoft extended their retirement by 24 months, to September 30, 2028. Custom tests sent from code with TrackAvailability() now sit in the archived Classic API guidance.

Where it comes up

Certificate expiration monitoring with Application Insights availability tests

Guide · Use Standard availability tests to catch expiring TLS certificates, set the lifetime check and alerts, price it, and when to use TrackAvailability instead.

Azure Monitor alerts explained: log alerts, action groups and the alert email

Guide · What an Azure Monitor alert email means, metric vs log search alerts, action groups, a log alert over KQL step by step, what alerts cost, and how to stop noise.

Availability results by test and location

KQL · One row per availability test and test location with the number of runs, the failed runs, the availability percentage and the p95 test duration in milliseconds, worst first.

Availability test failures with their messages

KQL · Up to 100 failed availability test runs from the last 24 hours, newest first, each with the test name, location, duration in milliseconds, failure message and operation_Id.

The practice and the tooling for measuring how a running application behaves for its users, chiefly response time, error rate and throughput, and tracing slow or failed work to its cause.

Percentile (p50, p95, p99)

The value a given share of measurements fall at or below; a p95 latency of 800 ms means 95% of requests took 800 ms or less, and one in twenty took longer.

Live Metrics

Application Insights' real-time view of incoming requests, failures, dependency calls and server performance, streamed from the app and never stored.

Where it comes up

Related terms

Sources, checked Oct 21, 2026