Azure环境下Java应用JMX MBeans对接Application Insights技术咨询

Great questions—let me walk you through what I know about both scenarios from working with Azure and Java apps:

问题1: Can Java Azure Functions expose metrics to JMX MBeans for Azure Monitor/Application Insights collection? #

The short answer is yes, it’s technically possible, but there are some caveats to keep in mind:

• Azure Functions’ Java runtime runs in a sandboxed environment, so JMX isn’t enabled by default. You’ll need to configure JVM startup parameters (like -Dcom.sun.management.jmxremote and related flags) to enable JMX access. These can be set via the JAVA_OPTS application setting in your Function App.
• You’ll also need the Application Insights Java Agent deployed alongside your Function App. The agent can be configured to scrape custom JMX MBeans by adding a settings.json file that defines which MBean objects and attributes to collect.
• Keep in mind Azure Functions’ event-driven lifecycle: instances spin up/down dynamically. Make sure your JMX metrics are tied to the function execution context correctly, so metrics don’t get lost during scaling.

That said, a more straightforward approach for Azure Functions is to use the built-in custom metrics API (via the TelemetryClient class) to send metrics directly to Application Insights. But if you specifically need to route metrics through JMX, the above steps will get you there.

问题2: Can Azure Monitor/Application Insights collect JMX metrics from a containerized third-party COTS Java app? #

Absolutely—and this is a common use case. There are a few reliable ways to do this:

Option 1: Use the Application Insights Java Agent (Recommended for COTS Apps) #

This is the most low-effort, non-intrusive method since you don’t need to modify the COTS app’s code:

• Package the Application Insights Java Agent JAR into your container image, or mount it as a volume.
• Update the container’s startup command to attach the agent: java -javaagent:/path/to/applicationinsights-agent.jar -jar your-cots-app.jar
• Configure the agent with a settings.json file that specifies the JMX MBeans you want to collect (define object names, attributes, and metric names).
• Set the APPLICATIONINSIGHTS_CONNECTION_STRING environment variable in the container to point to your Application Insights resource.

Option 2: Use JMX Exporter + Azure Monitor Container Insights #

If you prefer using Prometheus-style metrics:

• Deploy the jmx_exporter alongside your COTS app in the container (as a sidecar or part of the same pod).
• Configure the exporter to scrape your app’s JMX metrics and expose them in Prometheus format.
• Enable Azure Monitor Container Insights, which can natively scrape Prometheus metrics from containers and send them to Azure Monitor.

Both methods work well—go with the Java Agent if you want tight integration with Application Insights’ full telemetry suite, or the exporter route if you already use Prometheus in your stack.

内容的提问来源于stack exchange，提问作者RikSaunderson