Crate opentelemetry_jaeger

Expand description

Collects OpenTelemetry spans and reports them to a given Jaeger agent or collector endpoint, propagate the tracing context between the applications using Jaeger propagation format.

See the Jaeger Docs for details about Jaeger and deployment information.

Compiler support: requires rustc 1.56+

Quickstart

First make sure you have a running version of the Jaeger instance you want to send data to:

$ docker run -d -p6831:6831/udp -p6832:6832/udp -p16686:16686 -p14268:14268 jaegertracing/all-in-one:latest

Then install a new jaeger pipeline with the recommended defaults to start exporting telemetry:

use opentelemetry::trace::Tracer;
use opentelemetry::global;

#[tokio::main]
async fn main() -> Result<(), opentelemetry::trace::TraceError> {
    global::set_text_map_propagator(opentelemetry_jaeger::Propagator::new());
    let tracer = opentelemetry_jaeger::new_agent_pipeline().install_simple()?;

    tracer.in_span("doing_work", |cx| {
        // Traced app logic here...
    });

    global::shutdown_tracer_provider(); // export remaining spans

    Ok(())
}

Or if you are running on an async runtime like Tokio and want to report spans in batches

use opentelemetry::trace::Tracer;
use opentelemetry::global;
use opentelemetry::runtime::Tokio;

fn main() -> Result<(), opentelemetry::trace::TraceError> {
    global::set_text_map_propagator(opentelemetry_jaeger::Propagator::new());
    let tracer = opentelemetry_jaeger::new_agent_pipeline().install_batch(Tokio)?;

    tracer.in_span("doing_work", |cx| {
        // Traced app logic here...
    });

    global::shutdown_tracer_provider(); // export remaining spans

    Ok(())
}

For optimal performance, a batch exporter is recommended as the simple exporter will export each span synchronously on drop. You can enable the rt-tokio, rt-tokio-current-thread or rt-async-std features and specify a runtime on the pipeline builder to have a batch exporter configured for you automatically.

[dependencies]
opentelemetry = { version = "*", features = ["rt-tokio"] }
opentelemetry-jaeger = { version = "*", features = ["rt-tokio"] }

let tracer = opentelemetry_jaeger::new_agent_pipeline()
    .install_batch(opentelemetry::runtime::Tokio)?;

Jaeger Exporter From Environment Variables

The jaeger pipeline builder can be configured dynamically via environment variables. All variables are optional, a full list of accepted options can be found in the jaeger variables spec.

Jaeger Collector Example

If you want to skip the agent and submit spans directly to a Jaeger collector, you can enable the optional collector_client feature for this crate. This example expects a Jaeger collector running on http://localhost:14268.

[dependencies]
opentelemetry-jaeger = { version = "..", features = ["collector_client", "isahc_collector_client"] }

Then you can use the with_endpoint method to specify the endpoint:

// Note that this requires the `collector_client` feature.
// We enabled the `isahc_collector_client` feature for a default isahc http client.
// You can also provide your own implementation via .with_http_client() method.
use opentelemetry::trace::{Tracer, TraceError};

fn main() -> Result<(), TraceError> {
    let tracer = opentelemetry_jaeger::new_collector_pipeline()
        .with_endpoint("http://localhost:14268/api/traces")
        // optionally set username and password for authentication of the exporter.
        .with_username("username")
        .with_password("s3cr3t")
        .with_isahc()
        //.with_http_client(<your client>) provide custom http client implementation
        .install_batch(opentelemetry::runtime::Tokio)?;

    tracer.in_span("doing_work", |cx| {
        // Traced app logic here...
    });

    Ok(())
}

Resource, tags and service name

In order to export the spans in different format. opentelemetry uses its own model internally. Most of the jaeger spans’ concept can be found in this model. The full list of this mapping can be found in OpenTelemetry to Jaeger Transformation.

The process tags in jaeger spans will be mapped as resource in opentelemetry. You can set it through OTEL_RESOURCE_ATTRIBUTES environment variable or using with_trace_config.

Note that to avoid copying data multiple times. Jaeger exporter will uses resource stored in Exporter.

The tags in jaeger spans will be mapped as attributes in opentelemetry spans. You can set it through set_attribute method.

Each jaeger span requires a service name. This will be mapped as a resource with service.name key. You can set it using one of the following methods from highest priority to lowest priority.

with_service_name.
include a service.name key value pairs when configure resource using with_trace_config.
set the service name as OTEL_SERVCE_NAME environment variable.
set the service.name attributes in OTEL_RESOURCE_ATTRIBUTES.
if the service name is not provided by the above method. unknown_service will be used.

Based on the service name, we update/append the service.name process tags in jaeger spans.

Kitchen Sink Full Configuration

Example showing how to override all configuration options. See the CollectorPipeline and AgentPipeline docs for details of each option.

Export to agents

use opentelemetry::{sdk::{trace::{self, RandomIdGenerator, Sampler}, Resource}, global, KeyValue, trace::{Tracer, TraceError}};

fn main() -> Result<(), TraceError> {
    global::set_text_map_propagator(opentelemetry_jaeger::Propagator::new());
    let tracer = opentelemetry_jaeger::new_agent_pipeline()
        .with_endpoint("localhost:6831")
        .with_service_name("my_app")
        .with_max_packet_size(9_216)
        .with_auto_split_batch(true)
        .with_instrumentation_library_tags(false)
        .with_trace_config(
            trace::config()
                .with_sampler(Sampler::AlwaysOn)
                .with_id_generator(RandomIdGenerator::default())
                .with_max_events_per_span(64)
                .with_max_attributes_per_span(16)
                .with_max_events_per_span(16)
                 // resources will translated to tags in jaeger spans
                .with_resource(Resource::new(vec![KeyValue::new("key", "value"),
                          KeyValue::new("process_key", "process_value")])),
        )
        .install_batch(opentelemetry::runtime::Tokio)?;

    tracer.in_span("doing_work", |cx| {
        // Traced app logic here...
    });

    // export remaining spans. It's optional if you can accept spans loss for the last batch.
    global::shutdown_tracer_provider();

    Ok(())
}

Export to collectors

Note that this example requires collecotr_client and isahc_collector_client feature.

use opentelemetry::{sdk::{trace::{self, RandomIdGenerator, Sampler}, Resource}, global, KeyValue, trace::{Tracer, TraceError}};

fn main() -> Result<(), TraceError> {
    global::set_text_map_propagator(opentelemetry_jaeger::Propagator::new());
    let tracer = opentelemetry_jaeger::new_collector_pipeline()
        .with_endpoint("http://localhost:14250/api/trace") // set collector endpoint
        .with_service_name("my_app") // the name of the application
        .with_trace_config(
            trace::config()
                .with_sampler(Sampler::AlwaysOn)
                .with_id_generator(RandomIdGenerator::default())
                .with_max_events_per_span(64)
                .with_max_attributes_per_span(16)
                .with_max_events_per_span(16)
                // resources will translated to tags in jaeger spans
                .with_resource(Resource::new(vec![KeyValue::new("key", "value"),
                          KeyValue::new("process_key", "process_value")])),
        )
        // we config a surf http client with 2 seconds timeout
        // and have basic authentication header with username=username, password=s3cr3t
        .with_isahc() // requires `isahc_collector_client` feature
        .with_username("username")
        .with_password("s3cr3t")
        .with_timeout(std::time::Duration::from_secs(2))
        .install_batch(opentelemetry::runtime::Tokio)?;

    tracer.in_span("doing_work", |cx| {
        // Traced app logic here...
    });

    // export remaining spans. It's optional if you can accept spans loss for the last batch.
    global::shutdown_tracer_provider();

    Ok(())
}

Crate Feature Flags

The following crate feature flags are available:

collector_client: Export span data directly to a Jaeger collector. User MUST provide the http client.
hyper_collector_client: Export span data with Jaeger collector backed by a hyper default http client.
surf_collector_client: Export span data with Jaeger collector backed by a surf default http client.
reqwest_collector_client: Export span data with Jaeger collector backed by a reqwest http client.
reqwest_blocking_collector_client: Export span data with Jaeger collector backed by a reqwest blocking http client.
isahc_collector_client: Export span data with Jaeger collector backed by a isahc http client.
wasm_collector_client: Enable collector in wasm.

Support for recording and exporting telemetry asynchronously can be added via the following flags, it extends the opentelemetry feature:

rt-tokio: Enable sending UDP packets to Jaeger agent asynchronously when the tokio Multi-Threaded Scheduler is used.
rt-tokio-current-thread: Enable sending UDP packets to Jaeger agent asynchronously when the tokio Current-Thread Scheduler is used.
rt-async-std: Enable sending UDP packets to Jaeger agent asynchronously when the async-std runtime is used.

Supported Rust Versions

OpenTelemetry is built against the latest stable release. The minimum supported version is 1.56. The current OpenTelemetry version is not guaranteed to build on Rust versions earlier than the minimum supported version.

The current stable Rust compiler and the three most recent minor versions before it will always be supported. For example, if the current stable compiler version is 1.56, the minimum supported version will not be increased past 1.46, three minor versions prior. Increasing the minimum supported compiler version is not considered a semver breaking change as long as doing so complies with this policy.