Analyzing traces in the Jaeger interface

1. Searching #

Finding the right trace.

Criteria #

The scatter plot at the top of the results places each trace by date and duration: isolated points at the top are the slow requests to examine first.

2. Reading a trace #

Timeline and critical path.

The timeline #

• Each row is a span, indented under its parent; the bar shows its position and duration.
• A span in red is in error; its logs (events) often contain the exception.
• The critical path (highlighted) is the sequence of spans that determines the total duration: optimizing outside this path does not reduce latency.
• Staircase spans reveal sequential calls that could be parallelized; dozens of identical spans reveal an N+1 problem.

Good practice: First look for the longest « leaf » span (no children) on the critical path: that is where time is really spent.

Comparing two traces #

The Compare view overlays two traces (a fast one and a slow one) and colors spans present in only one or with different durations: ideal for understanding why the same operation is sometimes slow.

3. Summary views #

Beyond a single trace.

Dependencies and SPM #

Service Performance Monitoring relies on the OpenTelemetry Collector's spanmetrics connector, which turns spans into metrics stored in Prometheus; Jaeger queries them for the Monitor tab.

connectors:
  spanmetrics:
exporters:
  prometheus: { endpoint: 0.0.0.0:8889 }
service:
  pipelines:
    traces:
      receivers: [otlp]
      processors: [batch]
      exporters: [jaeger_storage_exporter, spanmetrics]
    metrics/spanmetrics:
      receivers: [spanmetrics]
      exporters: [prometheus]