Traces and the service browser

Traces and the trace viewer

A trace (route /traces/:traceId) is the waterfall of spans for one request across services. Open one from a service's endpoint, an error, a synthetic test, or a Workbook's APM Observe node.

Reading the waterfall (cookbook)

1. The root span at the top shows total duration — how slow the request was.
2. Find the longest span; it is usually the bottleneck (often a DB query or a downstream call).
3. Open a red span for its exception, status and attributes.
4. Watch for gaps between spans — queueing or uninstrumented work.
5. Use method-level profiling to jump from a slow span to the exact hot method.

Searching traces and the service browser

The service browser (/service-browser) lets you query and filter traces across services by attributes — service, endpoint, status, duration, custom tags — to build a list, then open any trace. Filter by status = error and duration > 1s to hunt slow failures.

Want this checked automatically? An APM Observe node in a Workbook asserts a service has no errors and meets a latency budget after a synthetic run.