Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Azure Kubernetes Service (AKS) autoinstrumentation is a feature that attaches Azure Monitor Application Insights software development kit (SDK) distros to your workload without requiring code changes. It works with workloads running as Kubernetes deployments in AKS. Currently, support for Java and Node.js is in public preview, while support for Python and .NET is in limited preview. This article details the onboarding process that allows a select group of limited preview customers to enable Python and .NET support for their AKS clusters.
Important
- This feature is a limited preview. Preview features are provided without a service-level agreement and aren't recommended for production workloads.
- For more information, see Supplemental Terms of Use for Microsoft Azure Previews.
You can use Python and .NET with both the AKS OpenTelemetry Protocol (OTLP) public preview and AKS autoinstrumentation (non-OTLP).
Onboard to the limited preview
This section outlines the process of enabling the limited preview feature. It significantly references public preview documentation at Autoinstrumentation for Azure Kubernetes Service (Preview), as most of the steps are the same. The documentation explicitly calls out limited preview-specific aspects.
Important
Application Insights experiences including prebuilt dashboards and queries expect and require OTLP metrics with delta temporality and exponential histogram aggregation.
When you use AKS autoinstrumentation, Azure Monitor automatically uses environment variables to configure SDKs to export metrics with delta temporality and exponential histograms. No additional user configuration is required.
For more information, see Metrics Exporters - OTLP.
- Review the Prerequisites section at Autoinstrumentation for Azure Kubernetes Service. The limitations apply to the limited preview except support is added for Python and .NET.
- Prepare the cluster by following the instructions at prepare a cluster.
- Choose a deployment that you want to instrument and onboard it by following the information at Per-deployment onboarding. The following points explain what you need to do differently for limited preview:
- Keep in mind that namespace-wide onboarding (described in the previous section) is not available for limited preview languages.
- Instead of using
public-previewannotations for Java and Node.js mentioned in the document (instrumentation.opentelemetry.io/inject-javaandinstrumentation.opentelemetry.io/inject-nodejs), usepublic-previewannotations:instrumentation.opentelemetry.io/private-preview-inject-pythonfor Pythoninstrumentation.opentelemetry.io/private-preview-inject-dotnetfor .NET
- If you don't have an Application Insights resource yet, create one. Then copy its connection string (found in the Overview area of the Application Insights resource) into the
spec.destination.applicationInsightsConnectionStringfield of the custom resource (CR), as instructed. - Put the CR into the same namespace as the deployment you're instrumenting.
- Place the
instrumentation.opentelemetry.io/private-preview-inject-*annotation correctly. Put it underspec.template.metadata.annotationsin the deployment, so that it exists at the pod level, not at the deployment level.
- Restart the deployment you're onboarding after setup completes, as described in the Restart deployment section. This step isn't required if the CR already exists when you add the annotation. Afterward, you don't need to restart or redeploy the deployment every time you change the CR.
- Make sure the deployment runs (under load if applicable). Wait three minutes and confirm that the Application Insights resource (or the underlying Log Analytics workspace) has telemetry.
Support
Contact us at otel@microsoft.com with your experiences, questions, or suggestions.