Non-official datadog tracing and log correlation for Rust services.
This crate contains the necessary glue to bridge the gap between OpenTelemetry and Datadog.
datadog-tracing has the following features:
1. tracing: utilities for building an OpenTelemetry tracer/layer that sends traces to the Datadog agent
2. log correlation: a log formatter that converts the trace ID and span ID to the Datadog native format and injects them into the dd.trace_id and dd.span_id fields
(more information)
3. propagation: a utility function to set the Datadog propagator as the global propagator
4. axum (enabled via the axum feature): re-exposing the functionality of axum-tracing-opentelemetry
5. opionated tracing-subscriber init function, configuring logs and the datadog exporter. It's optional, and you can build your own: the functions it uses are exposed.
The lib is configurable via environment variables as following:
| env var | default value | description |
|------------------------|---------------------------------------------|-----------------------------------------------------------|
| DDENABLED | false | Enables the datadog exporter and traceid/spanid on logs |
| DDSERVICE |
This lib was highly inspired on datadog-tracing, which is also a glue between tracing + opentelemtry + datadog. The main difference is that it exportes using the opentelemetryotlp exporter, and this one uses opentelemetrydatadog, so there is no need to configure your datadog agent to receive traces via OTLP and the default datadog APM works as expected!
Two commonly used propagation standards are B3 (OpenZipkin's propagation style)
and Jaeger. OpenTelemetry supports both.
Most Datadog SDK's support both B3 and the Datadog native propagation style.
For example, the Python datadog-tracing library supports B3 but it
needs to be explicitly enabled.
For ease of integration with services written in other languages that use the official Datadog SDK,
we opted for sticking with Datadog-style propagation over B3. This is set via the
set_global_propagator function which is automatically called when you create the tracer.
The Python library takes care of propagation of the trace context automatically. Unfortunately, we need to do this manually in Rust.
Arguably, propagation in HTTP requests is the most common need.
This crate does not provide any additional support, but we recommend using
the reqwest-middleware crate
to inject the necessary headers when using reqwest.
If you set the global propagator using datadog-tracing, it will work out of the box.
```rust use datadog-tracing::setglobalpropagator; use reqwestmiddleware::{ClientBuilder, ClientWithMiddleware}; use reqwesttracing::TracingMiddleware;
async fn main() { setglobalpropagator(); client = gethttpclient();
// configure tracing, setup your app and inject the client
}
fn gethttpclient() -> ClientWithMiddleware { ClientBuilder::new(reqwest::Client::new()) .with(TracingMiddleware::default()) .build() } ```