Instrumentation

Instrumentation for OpenTelemetry C++

You are viewing the English version of this page because it has not yet been fully translated. Interested in helping out? See Contributing.

插桩（Instrumentation）是指向应用中添加可观测性代码的行为。

如果你正在对一个应用进行插桩，需要使用适合你语言的 OpenTelemetry SDK。然后，你可以使用 SDK 初始化 OpenTelemetry，并使用 API 对代码进行插桩。这将从你的应用及其安装的任何带有插桩的库中导出遥测数据。

如果你正在对一个库进行插桩，只需安装适合你语言的 OpenTelemetry API 包。你的库不会自行导出遥测数据。只有当该库作为使用 OpenTelemetry SDK 的应用的一部分时，它才会导出遥测数据。有关如何对库进行插桩的更多信息，请参见 Libraries。

有关 OpenTelemetry API 和 SDK 的更多信息，请参见 specification。

Note

OpenTelemetry C++ doesn’t support automatic instrumentation when the source code of the library you want to instrument isn’t available.

Setup

Follow the instructions in the Getting Started Guide to build OpenTelemetry C++.

Traces

Initialize tracing

auto provider = opentelemetry::trace::Provider::GetTracerProvider();
auto tracer = provider->GetTracer("foo_library", "1.0.0");

The TracerProvider acquired in the first step is a singleton object that is usually provided by the OpenTelemetry C++ SDK. It is used to provide specific implementations for API interfaces. In case no SDK is used, the API provides a default no-op implementation of a TracerProvider.

The Tracer acquired in the second step is needed to create and start Spans.

Start a span

auto span = tracer->StartSpan("HandleRequest");

This creates a span, sets its name to "HandleRequest", and sets its start time to the current time. Refer to the API documentation for other operations that are available to enrich spans with additional data.

Mark a span as active

auto scope = tracer->WithActiveSpan(span);

This marks a span as active and returns a Scope object. The scope object controls how long a span is active. The span remains active for the lifetime of the scope object.

The concept of an active span is important, as any span that is created without explicitly specifying a parent is parented to the currently active span. A span without a parent is called root span.

Create nested spans

auto outer_span = tracer->StartSpan("Outer operation");
auto outer_scope = tracer->WithActiveSpan(outer_span);
{
    auto inner_span = tracer->StartSpan("Inner operation");
    auto inner_scope = tracer->WithActiveSpan(inner_span);
    // ... perform inner operation
    inner_span->End();
}
// ... perform outer operation
outer_span->End();

Spans can be nested, and have a parent-child relationship with other spans. When a given span is active, the newly created span inherits the active span’s trace ID, and other context attributes.

Context propagation

// set global propagator
opentelemetry::context::propagation::GlobalTextMapPropagator::SetGlobalPropagator(
    nostd::shared_ptr<opentelemetry::context::propagation::TextMapPropagator>(
        new opentelemetry::trace::propagation::HttpTraceContext()));

// get global propagator
HttpTextMapCarrier<opentelemetry::ext::http::client::Headers> carrier;
auto propagator =
    opentelemetry::context::propagation::GlobalTextMapPropagator::GetGlobalPropagator();

//inject context to headers
auto current_ctx = opentelemetry::context::RuntimeContext::GetCurrent();
propagator->Inject(carrier, current_ctx);

//Extract headers to context
auto current_ctx = opentelemetry::context::RuntimeContext::GetCurrent();
auto new_context = propagator->Extract(carrier, current_ctx);
auto remote_span = opentelemetry::trace::propagation::GetSpan(new_context);

Context contains the metadata of the currently active Span including Span ID, Trace ID, and flags. Context Propagation is an important mechanism in distributed tracing to transfer this Context across service boundary often through HTTP headers. OpenTelemetry provides a text-based approach to propagate context to remote services using the W3C Trace Context HTTP headers.

Metrics

Initialize exporter and reader

Initialize an exporter and a reader. In this case, you initialize an OStream Exporter which prints to stdout by default. The reader periodically collects metrics from the Aggregation Store and exports them.

std::unique_ptr<opentelemetry::sdk::metrics::MetricExporter> exporter{new opentelemetry::exporters::OStreamMetricExporter};
std::unique_ptr<opentelemetry::sdk::metrics::MetricReader> reader{
    new opentelemetry::sdk::metrics::PeriodicExportingMetricReader(std::move(exporter), options)};

Initialize a meter provider

Initialize a MeterProvider and add the reader. Use this to obtain Meter objects in the future.

auto provider = std::shared_ptr<opentelemetry::metrics::MeterProvider>(new opentelemetry::sdk::metrics::MeterProvider());
auto p = std::static_pointer_cast<opentelemetry::sdk::metrics::MeterProvider>(provider);
p->AddMetricReader(std::move(reader));

Create a counter

Create a Counter instrument from the Meter, and record the measurement. Every Meter pointer returned by the MeterProvider points to the same Meter. This means that the Meter can combine metrics captured from different functions without having to constantly pass the Meter around the library.

auto meter = provider->GetMeter(name, "1.2.0");
auto double_counter = meter->CreateDoubleCounter(counter_name);
// Create a label set which annotates metric values
std::map<std::string, std::string> labels = {{"key", "value"}};
auto labelkv = common::KeyValueIterableView<decltype(labels)>{labels};
double_counter->Add(val, labelkv);

Create a histogram

Create a histogram instrument from the meter, and record the measurement.

auto meter = provider->GetMeter(name, "1.2.0");
auto histogram_counter = meter->CreateDoubleHistogram("histogram_name");
histogram_counter->Record(val, labelkv);

Create an observable counter

Create an observable counter instrument from the meter, and add a callback. The callback is used to record the measurement during metrics collection. Ensure to keep the Instrument object active for the lifetime of collection.

auto meter = provider->GetMeter(name, "1.2.0");
auto counter = meter->CreateDoubleObservableCounter(counter_name);
counter->AddCallback(MeasurementFetcher::Fetcher, nullptr);

Create views

Map the counter instrument to sum aggregation

Create a view to map the Counter Instrument to Sum Aggregation. Add this view to provider. View creation is optional unless you want to add custom aggregation config, and attribute processor. Metrics SDK creates a missing view with default mapping between Instrument and Aggregation.

std::unique_ptr<opentelemetry::sdk::metrics::InstrumentSelector> instrument_selector{
    new opentelemetry::sdk::metrics::InstrumentSelector(opentelemetry::sdk::metrics::InstrumentType::kCounter, "counter_name")};
std::unique_ptr<opentelemetry::sdk::metrics::MeterSelector> meter_selector{
    new opentelemetry::sdk::metrics::MeterSelector(name, version, schema)};
std::unique_ptr<opentelemetry::sdk::metrics::View> sum_view{
    new opentelemetry::sdk::metrics::View{name, "description", opentelemetry::sdk::metrics::AggregationType::kSum}};
p->AddView(std::move(instrument_selector), std::move(meter_selector), std::move(sum_view));

Map the histogram instrument to histogram aggregation

std::unique_ptr<opentelemetry::sdk::metrics::InstrumentSelector> histogram_instrument_selector{
    new opentelemetry::sdk::metrics::InstrumentSelector(opentelemetry::sdk::metrics::InstrumentType::kHistogram, "histogram_name")};
std::unique_ptr<opentelemetry::sdk::metrics::MeterSelector> histogram_meter_selector{
    new opentelemetry::sdk::metrics::MeterSelector(name, version, schema)};
std::unique_ptr<opentelemetry::sdk::metrics::View> histogram_view{
    new opentelemetry::sdk::metrics::View{name, "description", opentelemetry::sdk::metrics::AggregationType::kHistogram}};
p->AddView(std::move(histogram_instrument_selector), std::move(histogram_meter_selector),
    std::move(histogram_view));

Map the observable counter instrument to sum aggregation

std::unique_ptr<opentelemetry::sdk::metrics::InstrumentSelector> observable_instrument_selector{
    new opentelemetry::sdk::metrics::InstrumentSelector(opentelemetry::sdk::metrics::InstrumentType::kObservableCounter,
                                     "observable_counter_name")};
std::unique_ptr<opentelemetry::sdk::metrics::MeterSelector> observable_meter_selector{
  new opentelemetry::sdk::metrics::MeterSelector(name, version, schema)};
std::unique_ptr<opentelemetry::sdk::metrics::View> observable_sum_view{
  new opentelemetry::sdk::metrics::View{name, "description", opentelemetry::sdk::metrics::AggregationType::kSum}};
p->AddView(std::move(observable_instrument_selector), std::move(observable_meter_selector),
         std::move(observable_sum_view));

Logs

The documentation for the logs API & SDK is missing, you can help make it available by editing this page.

Next steps

You’ll also want to configure an appropriate exporter to export your telemetry data to one or more telemetry backends.

Feedback

Was this page helpful?

Thank you. Your feedback is appreciated!

Please let us know how we can improve this page. Your feedback is appreciated!

Last modified June 11, 2025: Alert cleanup (#7090) (c392c714)