Trace Graph
The Trace Graph shows one transaction end to end: every event, every consumer, every broker hop it touched, and how they connect.
Reading the graph
Each node is one event: its type, the consumer or source service that produced or handled it, the broker it moved through, its status, latency, and retry count. A dashed node marked "unobserved upstream" means a hop happened before any instrumented service picked it up. That's expected if only part of your system is instrumented so far.
Switch between a tree view (parent/child shape) and a lineage view (strict chronological order) with the view toggle, and use the zoom controls in the bottom corner to fit a large trace on screen.
Summary chips across the top give you an at-a-glance count of failed, dead-lettered, retrying, and successful events in this trace, plus total processing time and how many services were involved.
Clicking into a node
Selecting a node opens a drawer with three tabs:
- Investigate: the full event detail for that hop, plus a way to search related events or browse open incidents.
- Recover: check idempotency coverage for this event type.
- System: jump to that consumer's page, or open the DLQ Center if this hop is dead-lettered.
Coming soon Replaying dead-lettered events straight from the dashboard isn't available yet. For now, use the DLQ Center to see what failed and why, then fix the underlying issue in your consumer or producer and let it reprocess normally.