Trace: represents a single request's complete path through a distributed system. A trace ID propagates in headers so all spans (units of work) across all services can be correlated into one tree.
2 / 5
What is a span in tracing terminology?
Span: a single operation inside a trace (e.g., an HTTP call, a database query). Spans have a name, start timestamp, duration, status, and key-value attributes. Parent-child relationships form the trace tree.
3 / 5
How is a trace ID propagated across service boundaries?
Context propagation: each service reads the trace context from incoming headers, creates child spans linked to the parent, and forwards the context to downstream calls. Standard formats (W3C traceparent, B3) ensure interoperability.
4 / 5
What is sampling in distributed tracing and why is it needed?
Sampling: tracing every request at high traffic creates prohibitive overhead and storage. Head sampling decides at the start of a request; tail sampling records if something interesting (error, slow) happened, balancing cost and observability.
5 / 5
What is the OpenTelemetry project?
OpenTelemetry (OTel): a CNCF project providing APIs, SDKs, and a collector for unified observability signals (traces, metrics, logs). It decouples instrumentation from backends, letting you switch between Jaeger, Zipkin, Datadog, etc.