Logs & Tracing

Logs and traces let teams reconstruct what happened to a specific request and why the system behaved the way it did.

Structured logs

Logs should include request identifiers, environment, category, provider, and execution outcome so a single event can be correlated across systems. Structured fields matter more than verbose free-text output.

Trace boundaries

Tracing should capture the major transitions in the request lifecycle: ingress, routing decision, provider execution, fallback, and response. These spans make latency attribution much easier during incidents.

Correlation

Every log line and trace span should include enough shared identifiers for operators to move between telemetry systems without guesswork. Correlation is critical when a request crosses application and platform boundaries.

Retention strategy

Retention should reflect both debugging needs and privacy constraints. Teams should keep enough detail for incident review without storing sensitive request material longer than necessary.

Next step

Review Data & Privacy to understand the boundaries on stored request data.