Ask Miss O11y: Tracing Is for Async, Too

Understanding the complexities of tracing asynchronous and multi-request processes is essential for gaining insights into a system's behavior. While distributed tracing effectively handles synchronous requests by creating a root span for the external request and child spans for subsequent actions, asynchronous tasks require a different approach. One method is to propagate the context to the consumer and continue the trace asynchronously, though this may lead to issues if spans fall outside the query's time window. Alternatively, span links can be used, allowing each consumer to create its own root span and trace, which can be linked back to the original trace to maintain independence while providing connectivity. This strategy accommodates multi-request processes by having a top-level trace and separate linked traces for individual requests. Despite OpenTelemetry tooling still maturing, Honeycomb offers tools to navigate these complexities, enabling developers to create meaningful stories from their code to better understand system processes.