See where the time goes.
A 4-second request: is it the database, the third-party API, or the fraud check? Stop guessing. The waterfall lays out every span on a shared timeline so the slow hop is obvious - its duration and status sit right next to it.
- Trace ingest through OTLP
- Trace list with service and time-range filters
- Span waterfall detail view
- Span duration and status inspection
- Trace percentile latency (p95, p99)
Pivot between the graph and the trace.
Latency rarely lives in one service. Click a slow edge on the Service Map, jump into a representative trace, take the numbers with you into a dashboard or an alert. No copy-paste between tabs, no "I'll find that trace ID in a minute."
- Sample trace drilldown from Service Map edges
- Trace count and trace p95 dashboard widgets
- CSV trace export on paid plans
Use it alongside.
Questions about traces.
Do you support OpenTelemetry trace context propagation?
Yes. W3C traceparent and baggage headers are honored on ingest, so distributed traces across services stitch together correctly.
What latency percentiles do you chart?
p50, p75, p90, p95, and p99 for any service-operation pair, with click-through from the chart to the trace list filtered to that latency band.
Do you sample traces?
No head sampling on ingest. You control sampling in your SDK or in the Collector via probabilistic or tail-based samplers - we store what arrives.
Can I filter spans by attribute?
Yes. Filter by service, operation, status, duration, and any span attribute. The waterfall view inspects attributes per span without leaving the trace.
Can I see logs from inside a trace?
Yes. Any log carrying the trace_id and span_id is linked from the span detail panel, so you can switch between waterfall and log context in one click.
See your first span in 60 seconds.
Point any OTLP tracer at Telemtra and open the waterfall.