Introduction to Infinite Tracing

Infinite Tracing is a fully managed cloud-based solution that can analyze 100% of your trace data and choose the most actionable data, letting you investigate and solve issues quickly.

This document only applies to the Infinite Tracing feature. If you are looking for our standard distributed tracing solution, see the standard distributed tracing setup guide.

What is Infinite Tracing?

Infinite Tracing allows you to send all your trace data to our cloud-based service and rely on Infinite Tracing to make sampling decisions. You can configure Infinite Tracing in various ways to ensure it's keeping the trace data you need to see.

Infinite Tracing requires you to send your telemetry to a trace observer, a dedicated, regional ingestion endpoint that buffers your spans in the cloud. The trace observer is the entry point for all Infinite Tracing data, and it accepts telemetry from the following sources:

• New Relic APM agents, via gRPC
• OpenTelemetry collectors and SDKs, via OTLP over gRPC or HTTP
• Other tools, such as Zipkin, via the Trace API

Unlike standard distributed tracing, Infinite Tracing can process more trace data. It uses tail-based sampling (sampling after data is collected), instead of the head-based sampling that standard tracing uses. Because Infinite Tracing makes the sampling decision after all spans in a trace arrive, it can keep the traces you need for troubleshooting—such as traces with errors or unusually slow traces—instead of deciding at the start of a request, before you know whether it will fail or be slow. Infinite Tracing also keeps a random sample of all traces, 1% by default, so you can see how your services behave during normal operation, not just when something goes wrong. You can adjust this value with the random trace filter, and fine-tune the error rules with the span attribute trace filter.

Resources for learning more about Infinite Tracing:

• Technical details of Infinite Tracing sampling and architecture

Requirements

Requirements differ depending on your pricing model:

• Usage-based pricing model: requires Pro or Enterprise edition.
• Original pricing model: requires New Relic help to enable it for your organization. For questions, contact your New Relic account representative.

The ability to set up this feature is governed by Infinite Tracing permissions.

Enable Infinite Tracing

When enabling Infinite Tracing, you should ideally enable it for all associated services. If some services in a trace use Infinite Tracing and others use our standard tracing solutions, each service makes its own sampling decision: standard tracing samples at the start of a request, while Infinite Tracing waits for the complete trace. As a result, spans from standard tracing services may be missing, which causes fragmented traces, orphaned spans, and inaccurate trace metrics. For details, see trace configuration conflicts.

To get started, follow the trace observer setup guide.

Configure Infinite Tracing

After enabling Infinite Tracing, you can configure trace observer filtering to ensure it retains the exact data you need.