Instrumentation Types: Automatic vs Manual
What you'll learn: How automatic and manual instrumentation differ in OpenTelemetry and when to use each approach.
Two Paths to Observability
OpenTelemetry offers two ways to add tracing to your applications: automatic instrumentation using language-specific agents, and manual instrumentation using the OTel API directly.
Automatic Instrumentation
Think of automatic instrumentation like installing cruise control in your car—it handles the common scenarios without you touching the steering wheel. Language-specific agents (like the Java agent or Python auto-instrumentation) inject tracing code into your application at runtime.
Benefits:
- Zero code changes required
- Instant visibility into frameworks and libraries (HTTP servers, database clients, message queues)
- Great for getting started quickly
Limitations:
- Only instruments known libraries and patterns
- Limited control over span names and attributes
- May not capture business-specific context
Manual Instrumentation
Manual instrumentation is like driving with full control—you decide exactly what to trace. You use the OpenTelemetry API to create spans, add attributes, and propagate context explicitly in your code.
Benefits:
- Fine-grained control over what gets traced
- Can add business-specific attributes (user IDs, transaction types, custom metrics)
- Trace internal functions and business logic
Limitations:
- Requires code changes
- More development effort
- Must maintain instrumentation as code evolves
The Hybrid Approach
Most production systems use both: automatic instrumentation for infrastructure-level visibility (HTTP, databases) and manual instrumentation for critical business flows and custom context.
Key Takeaway: Automatic instrumentation gives you quick, broad coverage with language agents; manual instrumentation provides precise, business-aware tracing through the OTel API—use both together for comprehensive observability.