OpenTelemetry Best Practices #2: Agents, Sidecars, Collectors, Coded Instrumentation

OpenTelemetry offers a modern approach to telemetry collection by enabling automatic and coded instrumentation, providing flexibility in how developers integrate it into their applications. Unlike traditional agent-based methods, OpenTelemetry's auto-instrumentation allows for hands-off integration without altering code, offering immediate insights similar to legacy APM tools but without being tied to specific backend platforms. However, while auto-instrumentation is a useful starting point for gathering baseline data, it lacks the context needed for more nuanced performance questions, which is where coded and custom instrumentation come into play. Coded instrumentation involves minimal code additions and is more naturally integrated into some languages, like .NET, while custom instrumentation allows teams to add specific context to their telemetry data, fostering a culture of observability. This approach empowers engineers to enhance their application's telemetry, making it easier to support production environments confidently.