Seeking Method to Monitor Custom Code with OpenTelemetry Auto-Instrumentation

[ansdudgh98]

Hello,

I am currently conducting a Proof of Concept (POC) to enhance our application's internal monitoring using OpenTelemetry.

📝 Environment
Platform: Self-managed Kubernetes Cluster

Instrumentation: OpenTelemetry Operator (using Instrumentation CRD for automatic instrumentation)

Languages: Java

Current Understanding & Challenge
My understanding so far is that OTel auto-instrumentation is fantastic for observing frameworks and libraries supported by the default SDK (e.g., Spring Boot, Flask, Django, JDBC, etc.).

However, it appears that our custom-implemented classes and specific business logic methods are not observed, which requires manual instrumentation (i.e., modifying the source code to add SDK calls).

My primary goal is to find a way to observe our custom-written code (e.g., specific methods or classes) while still leveraging the power and convenience of auto-instrumentation via the OTel Operator, for Java

I discovered a GitHub issue that perfectly describes a part of my problem: #2574

Has there been any progress, update, or official recommendation regarding that issue?

Aside from that specific issue, is there another recommended way (perhaps via Instrumentation CRD configuration) to achieve package-level or class-level instrumentation without modifying the application code?

[laurit]

If you wish to create spans for your methods then you can achieve it by adding the @WithSpan annotation to the method or using otel apis to manually instrument your code. Or use https://github.com/open-telemetry/opentelemetry-java-instrumentation/tree/main/instrumentation/methods If this is not enough for your use case then you can write your own extension https://github.com/open-telemetry/opentelemetry-java-instrumentation/tree/main/examples/extension Often creating a span for each method isn't reasonable. What you are most likely looking for is called profiling. Otel has plans for it but we don't have a working solution yet.