Making observability a first-class platform concern

OpenTelemetry has become the standard for vendor-neutral instrumentation, but running it well in Kubernetes requires more than enabling auto-instrumentation. Platform teams must think about collector topology, pipeline design, resource limits, configuration management, and how telemetry behaves under real production conditions.

In this workshop, we'll use a small microservices application as a controlled environment to examine how OpenTelemetry works end to end, from SDK instrumentation to Collector pipelines and Kubernetes operator-based deployment. Rather than focusing only on setup, we'll analyze architectural decisions: why use a Collector instead of direct export, how to structure pipelines, how operator-based instrumentation scales across namespaces, and what failure modes to expect in distributed systems.

By the end, participants will not just have a working setup, but a clearer mental model of how to design, evaluate, and operate OpenTelemetry in production-oriented Kubernetes environments.