Summary

In this chapter, we discussed suggestions and recommendations for telemetry collection. To describe some scenario or operation, we usually need multiple signals: tracing enables correlation and causation, logs provide additional information not covered by traces, events collect usage information, and metrics optimize instrumentations, queries, and alerts.

Depending on your application’s needs and stability, you can control costs by tuning the sampling rate on tracing and using metrics for performance data and events for usage reports.

OpenTelemetry semantic conventions provide instrumentation recipes for common technologies and concepts. By following them, you can create high-quality instrumentations with good defaults that you can tune to your needs. Observability backends can provide their best experiences to help you visualize, detect anomalies, and perform other semi-automated analyses. For proprietary technologies or application-specific instrumentation, where...