libdd-trace-stats
Compute aggregated statistics from distributed tracing spans with time-bucketed concentration.
Overview
libdd-trace-stats provides utilities for computing trace statistics by aggregating spans into time-based buckets with support for DDSketch distributions.
Features
- Span Concentration: Aggregate spans into time buckets
- DDSketch Integration: Use DDSketch for latency distribution metrics
- Peer Tags: Support for peer service aggregation
- Span Filtering: Filter spans by top-level, measured, or span.kind
- Time Bucketing: Configurable bucket sizes for aggregation
- Statistics Export: Generate statistics payloads for Datadog backend
- Stats Exporter (non-wasm): Periodic HTTP flusher backed by any
HttpClientTraitimplementation
Span Concentrator
The SpanConcentrator is the core component that aggregates spans into statistics:
Aggregation
Spans are aggregated into time buckets based on their end time. Within each bucket, spans are further aggregated by:
- Service name
- Resource name
- Operation name
- Span type
- HTTP status code
- Peer tags (if enabled)
Span Eligibility
Only certain spans are aggregated:
- Root spans
- Top-level spans
- Measured spans
- Spans with eligible
span.kindvalues
Flushing
When flushed, the concentrator keeps the most recent buckets and returns older buckets as statistics.
Stats Exporter
StatsExporter wraps a FlushableConcentrator and periodically drains it, encoding the resulting ClientStatsPayload as msgpack and POSTing it to the agent's /v0.6/stats endpoint.
use ;
use NativeCapabilities;
use ;
use Duration;
let exporter = new;
// Flush and send (force=true drains all buckets on shutdown)
exporter.send.await?;
Example Usage
use SpanConcentrator;
use ;
// Create a concentrator
let mut concentrator = new;
// Add spans
// concentrator.add_span(&span);
// Flush statistics
// let stats = concentrator.flush(false);