Why architecture matters here

Traces fall apart at batch boundaries without span links. Architecture matters because parent + links + batch compose.

Advertisement

The architecture: every piece explained

The top strip is basics. Parent span. Follows_from link. Batch consumer. Link attributes.

The middle row is pipeline. Producer traces. Consumer span. UI traversal. Sampling.

The lower rows are ops. Cost. Metrics. Ops — schema + backend support + retention.

Span links — follows_from + batch correlation + multi-traceconnect spans across tracesParent spansame traceFollows_from linkcross-trace refBatch consumerN msgs -> 1 spanLink attributesreason + contextProducer tracesmany separateConsumer spanwith N linksUI traversalclick link to jumpSamplinglink-awareCostlink count per spanMetricslink cardinalityOps — schema + backend support + retentionconnectcollecttraversesamplebudgetwatchwatchoperateoperate
Span links connect a consumer to many producer traces.
Advertisement

End-to-end flow

End-to-end: 100 producers each start a trace and enqueue Kafka messages with traceparent header. Consumer batches 50 messages, starts one span with 50 follows_from links. UI shows consumer span and lets user pivot to any of the 50 traces.