Decide what the records need to explain
Business reviewers need figures and definitions. Operations staff need the failed stage and elapsed time. Security reviewers need identities, resources and access decisions. These tasks can share a request identifier while collecting different levels of detail.
Decide who may read the records, their retention period and when they will be archived or deleted before choosing fields. Use synthetic content for public demonstrations and feedback. Omit real credentials and unrelated business records.
Design the smallest sufficient record
| Record dimension | Suggested fields |
|---|---|
| Request identification | Request ID, time, deployment version and configuration version |
| Access | Appropriately protected identity identifier, access scope and decision |
| Execution | Stage, status, elapsed time and relevant error category |
| Evidence | Data snapshot, metric definition and document version |
| Delivery | Report status and controlled file location |
Correlate the stages using one identifier
Use the request ID to follow an interaction from its entry point through query or retrieval, result validation and export. OpenTelemetry distinguishes metrics, logs and traces. In diagnosis, aggregate trends can reveal an anomaly and individual records can help explain its cause.
These fields are design suggestions. Compare them with what the actual deployment emits and verify its time zones, identifier correlation and failure paths before depending on a particular record format.
Test traceability
- Run a question with a known answer, then recover the data scope and citations using only the records.
- Introduce a controlled connection failure and check that authentication, network and query stages can be distinguished.
- Confirm that ordinary users cannot read audit records outside their scope.
- At the retention deadline, verify that deletion or archiving also covers copies and exported files.
