Trace Groups

이 페이지는 아직 영어로 제공되지 않습니다. 번역 작업 중입니다.
현재 번역 프로젝트에 대한 질문이나 피드백이 있으신 경우 언제든지 연락주시기 바랍니다.

Request access to the Preview!

Trace Groups are in Preview. To request access, contact Datadog support.

Overview

Trace groups allow you to define custom grouping keys for your trace data, enabling you to discover request count trends, performance bottlenecks, and error hotspots by looking at aggregated data instead of individual spans and traces.

From the Trace Explorer query editor, you can group your results by any span tag or attribute to observe:

Request counts
Error rates
Latency distributions

These metrics are organized within each group based on the distinct values of your chosen grouping keys.

Query editor

Navigate to the Trace Explorer.
Enter your query in the search bar.
Select up to four dimensions in the Group by clause, using any span tag or attribute.

After selecting a dimension to group by, you can specify where to get the dimension’s values from using the from dropdown:

Span: Group by the dimension of the queried span (default). For example, a.
Parent of span: Group by the specified dimension from the parent span of spans matching the query. For example, to visualize how an API endpoint performs based on the service calling it, group by service from parent(a).
Root span: Group by the specified dimension from the root span of the trace. For example, to analyze backend request patterns based on the frontend pages requests originate from, group by @view.name from root.

View trace groups in the group list

Trace groups are displayed as unique values of the selected dimension. Each group is shown with three key metrics:

REQUESTS: Count of spans within the group.
ERRORS: Error rate and count of errors.
P95 Latency: p95 latency of spans.

To view these metrics aggregated over the parent or root span instead of the queried span, select parent(a) or root in the Show metrics from statement.

Additionally, the Latency Breakdown surfaces how time is spent between different services within requests from each group, allowing you to visually spot latency bottlenecks for given groups.