Manage Telemetry with SDK

The SDK is the built-in reference implementation of the API, processing and exporting telemetry produced by instrumentation API calls. This page is a conceptual overview of the SDK, including descriptions, links to relevant Javadocs, artifact coordinates, sample programmatic configurations and more. See Configure the SDK for details on SDK configuration, including zero-code SDK autoconfigure.

The SDK consists of the following top level components:

SdkTracerProvider: The SDK implementation of TracerProvider, including tools for sampling, processing, and exporting spans.
SdkMeterProvider: The SDK implementation of MeterProvider, including tools for configuring metric streams and reading / exporting metrics.
SdkLoggerProvider: The SDK implementation of LoggerProvider, including tools for processing and exporting logs.
TextMapPropagator: Propagates context across process boundaries.

These are combined into OpenTelemetrySdk, a carrier object which makes it convenient to pass fully-configured SDK components to instrumentation.

The SDK comes packaged with a variety of built-in components which are sufficient for many use cases, and supports plugin interfaces for extensibility.

SDK plugin extension interfaces

When built-in components are insufficient, the SDK can be extended by implementing various plugin extension interfaces:

Sampler: Configures which spans are recorded and sampled.
SpanProcessor: Processes spans when they start and end.
SpanExporter: Exports spans out of process.
MetricReader: Reads aggregated metrics.
MetricExporter: Exports metrics out of process.
LogRecordProcessor: Processes log records when they are emitted.
LogRecordExporter: Exports log records out of process.
TextMapPropagator: Propagates context across process boundaries.

SDK components

The io.opentelemetry:opentelemetry-sdk:1.45.0 artifact contains the OpenTelemetry SDK.

The following sections describe the core user-facing components of the SDK. Each component section includes:

A brief description, including a link to the Javadoc type reference.
If the component is a plugin extension interface, a table of available built-in and opentelemetry-java-contrib implementations.
A simple demonstration of programmatic-configuration.
If the component is a plugin extension interface, a simple demonstration of a custom implementation.

OpenTelemetrySdk

OpenTelemetrySdk is the SDK implementation of OpenTelemetry. It is a holder for top-level SDK components which makes it convenient to pass fully-configured SDK components to instrumentation.

OpenTelemetrySdk is configured by the application owner, and consists of:

SdkTracerProvider: The SDK implementation of TracerProvider.
SdkMeterProvider: The SDK implementation of MeterProvider.
SdkLoggerProvider: The SDK implementation of LoggerProvider.
ContextPropagators: Propagates context across process boundaries.

The following code snippet demonstrates OpenTelemetrySdk programmatic configuration:

package otel;

import io.opentelemetry.sdk.OpenTelemetrySdk;
import io.opentelemetry.sdk.resources.Resource;

public class OpenTelemetrySdkConfig {
  public static OpenTelemetrySdk create() {
    Resource resource = ResourceConfig.create();
    return OpenTelemetrySdk.builder()
        .setTracerProvider(SdkTracerProviderConfig.create(resource))
        .setMeterProvider(SdkMeterProviderConfig.create(resource))
        .setLoggerProvider(SdkLoggerProviderConfig.create(resource))
        .setPropagators(ContextPropagatorsConfig.create())
        .build();
  }
}

Resource

Resource is a set of attributes defining the telemetry source. An application should associate the same resource with SdkTracerProvider, SdkMeterProvider, SdkLoggerProvider.

ResourceProviders contribute contextual information to the autoconfigured resource based on the environment. See documentation for list of available ResourceProviders.

The following code snippet demonstrates Resource programmatic configuration:

package otel;

import io.opentelemetry.sdk.resources.Resource;
import io.opentelemetry.semconv.ServiceAttributes;

public class ResourceConfig {
  public static Resource create() {
    return Resource.getDefault().toBuilder()
        .put(ServiceAttributes.SERVICE_NAME, "my-service")
        .build();
  }
}

SdkTracerProvider

SdkTracerProvider is the SDK implementation of TracerProvider, and is responsible for handling trace telemetry produced by the API.

SdkTracerProvider is configured by the application owner, and consists of:

Resource: The resource spans are associated with.
Sampler: Configures which spans are recorded and sampled.
SpanProcessors: Processes spans when they start and end.
SpanExporters: Exports spans out of process (in conjunction with associated with SpanProcessors).
SpanLimits: Controls the limits of data associated with spans.

The following code snippet demonstrates SdkTracerProvider programmatic configuration:

package otel;

import io.opentelemetry.sdk.resources.Resource;
import io.opentelemetry.sdk.trace.SdkTracerProvider;

public class SdkTracerProviderConfig {
  public static SdkTracerProvider create(Resource resource) {
    return SdkTracerProvider.builder()
        .setResource(resource)
        .addSpanProcessor(
            SpanProcessorConfig.batchSpanProcessor(
                SpanExporterConfig.otlpHttpSpanExporter("http://localhost:4318/v1/spans")))
        .setSampler(SamplerConfig.parentBasedSampler(SamplerConfig.traceIdRatioBased(.25)))
        .setSpanLimits(SpanLimitsConfig::spanLimits)
        .build();
  }
}

Sampler

A Sampler is a plugin extension interface responsible for determining which spans are recorded and sampled.

By default SdkTracerProvider is configured with the ParentBased(root=AlwaysOn) sampler. This results in 100% of spans being sampled if unless a calling application performs sampling. If this is too noisy / expensive, change the sampler.

Samplers built-in to the SDK and maintained by the community in opentelemetry-java-contrib:

Class	Artifact	Description
`ParentBased`	`io.opentelemetry:opentelemetry-sdk:1.45.0`	Samples spans based on sampling status of the span’s parent.
`AlwaysOn`	`io.opentelemetry:opentelemetry-sdk:1.45.0`	Samples all spans.
`AlwaysOff`	`io.opentelemetry:opentelemetry-sdk:1.45.0`	Drops all spans.
`TraceIdRatioBased`	`io.opentelemetry:opentelemetry-sdk:1.45.0`	Samples spans based on a configurable ratio.
`JaegerRemoteSampler`	`io.opentelemetry:opentelemetry-sdk-extension-jaeger-remote-sampler:1.45.0`	Samples spans based on configuration from a remote server.
`LinksBasedSampler`	`io.opentelemetry.contrib:opentelemetry-samplers:1.42.0-alpha`	Samples spans based on sampling status of the span’s links.
`RuleBasedRoutingSampler`	`io.opentelemetry.contrib:opentelemetry-samplers:1.42.0-alpha`	Samples spans based on configurable rules.
`ConsistentSamplers`	`io.opentelemetry.contrib:opentelemetry-consistent-sampling:1.42.0-alpha`	Various consistent sampler implementations as defined by probability sampling.

The following code snippet demonstrates Sampler programmatic configuration:

package otel;

import io.opentelemetry.sdk.extension.trace.jaeger.sampler.JaegerRemoteSampler;
import io.opentelemetry.sdk.trace.samplers.Sampler;
import java.time.Duration;

public class SamplerConfig {
  public static Sampler parentBasedSampler(Sampler root) {
    return Sampler.parentBasedBuilder(root)
        .setLocalParentNotSampled(Sampler.alwaysOff())
        .setLocalParentSampled(Sampler.alwaysOn())
        .setRemoteParentNotSampled(Sampler.alwaysOff())
        .setRemoteParentSampled(Sampler.alwaysOn())
        .build();
  }

  public static Sampler alwaysOn() {
    return Sampler.alwaysOn();
  }

  public static Sampler alwaysOff() {
    return Sampler.alwaysOff();
  }

  public static Sampler traceIdRatioBased(double ratio) {
    return Sampler.traceIdRatioBased(ratio);
  }

  public static Sampler jaegerRemoteSampler() {
    return JaegerRemoteSampler.builder()
        .setInitialSampler(Sampler.alwaysOn())
        .setEndpoint("http://endpoint")
        .setPollingInterval(Duration.ofSeconds(60))
        .setServiceName("my-service-name")
        .build();
  }
}

Implement the Sampler interface to provide your own custom sampling logic. For example:

package otel;

import io.opentelemetry.api.common.Attributes;
import io.opentelemetry.api.trace.SpanKind;
import io.opentelemetry.context.Context;
import io.opentelemetry.sdk.trace.data.LinkData;
import io.opentelemetry.sdk.trace.samplers.Sampler;
import io.opentelemetry.sdk.trace.samplers.SamplingResult;
import java.util.List;

public class CustomSampler implements Sampler {
  @Override
  public SamplingResult shouldSample(
      Context parentContext,
      String traceId,
      String name,
      SpanKind spanKind,
      Attributes attributes,
      List<LinkData> parentLinks) {
    // Callback invoked when span is started, before any SpanProcessor is called.
    // If the SamplingDecision is:
    // - DROP: the span is dropped. A valid span context is created and SpanProcessor#onStart is
    // still called, but no data is recorded and SpanProcessor#onEnd is not called.
    // - RECORD_ONLY: the span is recorded but not sampled. Data is recorded to the span,
    // SpanProcessor#onStart and SpanProcessor#onEnd are called, but the span's sampled status
    // indicates it should not be exported out of process.
    // - RECORD_AND_SAMPLE: the span is recorded and sampled. Data is recorded to the span,
    // SpanProcessor#onStart and SpanProcessor#onEnd are called, and the span's sampled status
    // indicates it should be exported out of process.
    return SpanKind.SERVER == spanKind ? SamplingResult.recordAndSample() : SamplingResult.drop();
  }

  @Override
  public String getDescription() {
    // Return a description of the sampler.
    return this.getClass().getSimpleName();
  }
}

SpanProcessor

A SpanProcessor is a plugin extension interface with callbacks invoked when a span is started and ended. They are often paired with SpanExporters to export spans out of process, but have other applications such as data enrichment.

Span processors built-in to the SDK and maintained by the community in opentelemetry-java-contrib:

Class	Artifact	Description
`BatchSpanProcessor`	`io.opentelemetry:opentelemetry-sdk:1.45.0`	Batches sampled spans and exports them via a configurable `SpanExporter`.
`SimpleSpanProcessor`	`io.opentelemetry:opentelemetry-sdk:1.45.0`	Exports each sampled span via a configurable `SpanExporter`.
`BaggageSpanProcessor`	`io.opentelemetry.contrib:opentelemetry-baggage-processor:1.42.0-alpha`	Enriches spans with baggage.
`JfrSpanProcessor`	`io.opentelemetry.contrib:opentelemetry-jfr-events:1.42.0-alpha`	Creates JFR events from spans.
`StackTraceSpanProcessor`	`io.opentelemetry.contrib:opentelemetry-span-stacktrace:1.42.0-alpha`	Enriches select spans with stack trace data.
`InferredSpansProcessor`	`io.opentelemetry.contrib:opentelemetry-inferred-spans:1.42.0-alpha`	Generates spans from async profiler instead of instrumentation.

The following code snippet demonstrates SpanProcessor programmatic configuration:

package otel;

import io.opentelemetry.sdk.trace.SpanProcessor;
import io.opentelemetry.sdk.trace.export.BatchSpanProcessor;
import io.opentelemetry.sdk.trace.export.SimpleSpanProcessor;
import io.opentelemetry.sdk.trace.export.SpanExporter;
import java.time.Duration;

public class SpanProcessorConfig {
  public static SpanProcessor batchSpanProcessor(SpanExporter spanExporter) {
    return BatchSpanProcessor.builder(spanExporter)
        .setMaxQueueSize(2048)
        .setExporterTimeout(Duration.ofSeconds(30))
        .setScheduleDelay(Duration.ofSeconds(5))
        .build();
  }

  public static SpanProcessor simpleSpanProcessor(SpanExporter spanExporter) {
    return SimpleSpanProcessor.builder(spanExporter).build();
  }
}

Implement the SpanProcessor interface to provide your own custom span processing logic. For example:

package otel;

import io.opentelemetry.context.Context;
import io.opentelemetry.sdk.common.CompletableResultCode;
import io.opentelemetry.sdk.trace.ReadWriteSpan;
import io.opentelemetry.sdk.trace.ReadableSpan;
import io.opentelemetry.sdk.trace.SpanProcessor;

public class CustomSpanProcessor implements SpanProcessor {

  @Override
  public void onStart(Context parentContext, ReadWriteSpan span) {
    // Callback invoked when span is started.
    // Enrich the record with a custom attribute.
    span.setAttribute("my.custom.attribute", "hello world");
  }

  @Override
  public boolean isStartRequired() {
    // Indicate if onStart should be called.
    return true;
  }

  @Override
  public void onEnd(ReadableSpan span) {
    // Callback invoked when span is ended.
  }

  @Override
  public boolean isEndRequired() {
    // Indicate if onEnd should be called.
    return false;
  }

  @Override
  public CompletableResultCode shutdown() {
    // Optionally shutdown the processor and cleanup any resources.
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public CompletableResultCode forceFlush() {
    // Optionally process any records which have been queued up but not yet processed.
    return CompletableResultCode.ofSuccess();
  }
}

SpanExporter

A SpanExporter is a plugin extension interface responsible for exporting spans out of process. Rather than directly registering with SdkTracerProvider, they are paired with SpanProcessors (typically BatchSpanProcessor).

Span exporters built-in to the SDK and maintained by the community in opentelemetry-java-contrib:

Class	Artifact	Description
`OtlpHttpSpanExporter` [1]	`io.opentelemetry:opentelemetry-exporter-otlp:1.45.0`	Exports spans via OTLP `http/protobuf`.
`OtlpGrpcSpanExporter` [1]	`io.opentelemetry:opentelemetry-exporter-otlp:1.45.0`	Exports spans via OTLP `grpc`.
`LoggingSpanExporter`	`io.opentelemetry:opentelemetry-exporter-logging:1.45.0`	Logs spans to JUL in a debugging format.
`OtlpJsonLoggingSpanExporter`	`io.opentelemetry:opentelemetry-exporter-logging-otlp:1.45.0`	Logs spans to JUL in an OTLP JSON encoding.
`OtlpStdoutSpanExporter`	`io.opentelemetry:opentelemetry-exporter-logging-otlp:1.45.0`	Logs spans to `System.out` in the OTLP JSON file encoding (experimental).
`ZipkinSpanExporter`	`io.opentelemetry:opentelemetry-exporter-zipkin:1.45.0`	Export spans to Zipkin.
`InterceptableSpanExporter`	`io.opentelemetry.contrib:opentelemetry-processors:1.42.0-alpha`	Passes spans to a flexible interceptor before exporting.
`KafkaSpanExporter`	`io.opentelemetry.contrib:opentelemetry-kafka-exporter:1.42.0-alpha`	Exports spans by writing to a Kafka topic.

[1]: See OTLP exporter sender for implementation details.

The following code snippet demonstrates SpanExporter programmatic configuration:

package otel;

import io.opentelemetry.exporter.logging.LoggingSpanExporter;
import io.opentelemetry.exporter.logging.otlp.OtlpJsonLoggingSpanExporter;
import io.opentelemetry.exporter.otlp.http.trace.OtlpHttpSpanExporter;
import io.opentelemetry.exporter.otlp.trace.OtlpGrpcSpanExporter;
import io.opentelemetry.sdk.trace.export.SpanExporter;
import java.time.Duration;

public class SpanExporterConfig {
  public static SpanExporter otlpHttpSpanExporter(String endpoint) {
    return OtlpHttpSpanExporter.builder()
        .setEndpoint(endpoint)
        .addHeader("api-key", "value")
        .setTimeout(Duration.ofSeconds(10))
        .build();
  }

  public static SpanExporter otlpGrpcSpanExporter(String endpoint) {
    return OtlpGrpcSpanExporter.builder()
        .setEndpoint(endpoint)
        .addHeader("api-key", "value")
        .setTimeout(Duration.ofSeconds(10))
        .build();
  }

  public static SpanExporter logginSpanExporter() {
    return LoggingSpanExporter.create();
  }

  public static SpanExporter otlpJsonLoggingSpanExporter() {
    return OtlpJsonLoggingSpanExporter.create();
  }
}

Implement the SpanExporter interface to provide your own custom span export logic. For example:

package otel;

import io.opentelemetry.sdk.common.CompletableResultCode;
import io.opentelemetry.sdk.trace.data.SpanData;
import io.opentelemetry.sdk.trace.export.SpanExporter;
import java.util.Collection;
import java.util.logging.Level;
import java.util.logging.Logger;

public class CustomSpanExporter implements SpanExporter {

  private static final Logger logger = Logger.getLogger(CustomSpanExporter.class.getName());

  @Override
  public CompletableResultCode export(Collection<SpanData> spans) {
    // Export the records. Typically, records are sent out of process via some network protocol, but
    // we simply log for illustrative purposes.
    logger.log(Level.INFO, "Exporting spans");
    spans.forEach(span -> logger.log(Level.INFO, "Span: " + span));
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public CompletableResultCode flush() {
    // Export any records which have been queued up but not yet exported.
    logger.log(Level.INFO, "flushing");
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public CompletableResultCode shutdown() {
    // Shutdown the exporter and cleanup any resources.
    logger.log(Level.INFO, "shutting down");
    return CompletableResultCode.ofSuccess();
  }
}

SpanLimits

SpanLimits defines constraints for the data captured by spans, including max attribute length, max number of attributes, and more.

The following code snippet demonstrates SpanLimits programmatic configuration:

package otel;

import io.opentelemetry.sdk.trace.SpanLimits;

public class SpanLimitsConfig {
  public static SpanLimits spanLimits() {
    return SpanLimits.builder()
        .setMaxNumberOfAttributes(128)
        .setMaxAttributeValueLength(1024)
        .setMaxNumberOfLinks(128)
        .setMaxNumberOfAttributesPerLink(128)
        .setMaxNumberOfEvents(128)
        .setMaxNumberOfAttributesPerEvent(128)
        .build();
  }
}

SdkMeterProvider

SdkMeterProvider is the SDK implementation of MeterProvider, and is responsible for handling metric telemetry produced by the API.

SdkMeterProvider is configured by the application owner, and consists of:

Resource: The resource metrics are associated with.
MetricReader: Reads the aggregated state of metrics.
- Optionally, with CardinalityLimitSelector for overriding cardinality limit by instrument kind. If unset, each instrument is limited to 2000 unique combinations of attributes per collection cycle. Cardinality limits are also configurable for individual instruments via views. See cardinality limits for more details.
MetricExporter: Exports metrics out of process (in conjunction with associated MetricReader).
Views: Configures metric streams, including dropping unused metrics.

The following code snippet demonstrates SdkMeterProvider programmatic configuration:

package otel;

import io.opentelemetry.sdk.metrics.SdkMeterProvider;
import io.opentelemetry.sdk.metrics.SdkMeterProviderBuilder;
import io.opentelemetry.sdk.resources.Resource;
import java.util.List;
import java.util.Set;

public class SdkMeterProviderConfig {
  public static SdkMeterProvider create(Resource resource) {
    SdkMeterProviderBuilder builder =
        SdkMeterProvider.builder()
            .setResource(resource)
            .registerMetricReader(
                MetricReaderConfig.periodicMetricReader(
                    MetricExporterConfig.otlpHttpMetricExporter(
                        "http://localhost:4318/v1/metrics")));
    // Uncomment to optionally register metric reader with cardinality limits
    // builder.registerMetricReader(
    //     MetricReaderConfig.periodicMetricReader(
    //         MetricExporterConfig.otlpHttpMetricExporter("http://localhost:4318/v1/metrics")),
    //     instrumentType -> 100);

    ViewConfig.dropMetricView(builder, "some.custom.metric");
    ViewConfig.histogramBucketBoundariesView(
        builder, "http.server.request.duration", List.of(1.0, 5.0, 10.0));
    ViewConfig.attributeFilterView(
        builder, "http.client.request.duration", Set.of("http.request.method"));
    ViewConfig.cardinalityLimitsView(builder, "http.server.active_requests", 100);
    return builder.build();
  }
}

MetricReader

A MetricReader is a plugin extension interface which is responsible for reading aggregated metrics. They are often paired with MetricExporters to export metrics out of process, but may also be used to serve the metrics to external scrapers in pull-based protocols.

Metric readers built-in to the SDK and maintained by the community in opentelemetry-java-contrib:

Class	Artifact	Description
`PeriodicMetricReader`	`io.opentelemetry:opentelemetry-sdk:1.45.0`	Reads metrics on a periodic basis and exports them via a configurable `MetricExporter`.
`PrometheusHttpServer`	`io.opentelemetry:opentelemetry-exporter-prometheus:1.45.0-alpha`	Serves metrics on an HTTP server in various prometheus formats.

The following code snippet demonstrates MetricReader programmatic configuration:

package otel;

import io.opentelemetry.exporter.prometheus.PrometheusHttpServer;
import io.opentelemetry.sdk.metrics.export.MetricExporter;
import io.opentelemetry.sdk.metrics.export.MetricReader;
import io.opentelemetry.sdk.metrics.export.PeriodicMetricReader;
import java.time.Duration;

public class MetricReaderConfig {
  public static MetricReader periodicMetricReader(MetricExporter metricExporter) {
    return PeriodicMetricReader.builder(metricExporter).setInterval(Duration.ofSeconds(60)).build();
  }

  public static MetricReader prometheusMetricReader() {
    return PrometheusHttpServer.builder().setHost("localhost").setPort(9464).build();
  }
}

Implement the MetricReader interface to provide your own custom metric reader logic. For example:

package otel;

import io.opentelemetry.sdk.common.CompletableResultCode;
import io.opentelemetry.sdk.common.export.MemoryMode;
import io.opentelemetry.sdk.metrics.Aggregation;
import io.opentelemetry.sdk.metrics.InstrumentType;
import io.opentelemetry.sdk.metrics.data.AggregationTemporality;
import io.opentelemetry.sdk.metrics.export.AggregationTemporalitySelector;
import io.opentelemetry.sdk.metrics.export.CollectionRegistration;
import io.opentelemetry.sdk.metrics.export.MetricReader;
import java.util.concurrent.Executors;
import java.util.concurrent.ScheduledExecutorService;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicReference;
import java.util.logging.Level;
import java.util.logging.Logger;

public class CustomMetricReader implements MetricReader {

  private static final Logger logger = Logger.getLogger(CustomMetricExporter.class.getName());

  private final ScheduledExecutorService executorService = Executors.newScheduledThreadPool(1);
  private final AtomicReference<CollectionRegistration> collectionRef =
      new AtomicReference<>(CollectionRegistration.noop());

  @Override
  public void register(CollectionRegistration collectionRegistration) {
    // Callback invoked when SdkMeterProvider is initialized, providing a handle to collect metrics.
    collectionRef.set(collectionRegistration);
    executorService.scheduleWithFixedDelay(this::collectMetrics, 0, 60, TimeUnit.SECONDS);
  }

  private void collectMetrics() {
    // Collect metrics. Typically, records are sent out of process via some network protocol, but we
    // simply log for illustrative purposes.
    logger.log(Level.INFO, "Collecting metrics");
    collectionRef
        .get()
        .collectAllMetrics()
        .forEach(metric -> logger.log(Level.INFO, "Metric: " + metric));
  }

  @Override
  public CompletableResultCode forceFlush() {
    // Export any records which have been queued up but not yet exported.
    logger.log(Level.INFO, "flushing");
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public CompletableResultCode shutdown() {
    // Shutdown the exporter and cleanup any resources.
    logger.log(Level.INFO, "shutting down");
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public AggregationTemporality getAggregationTemporality(InstrumentType instrumentType) {
    // Specify the required aggregation temporality as a function of instrument type
    return AggregationTemporalitySelector.deltaPreferred()
        .getAggregationTemporality(instrumentType);
  }

  @Override
  public MemoryMode getMemoryMode() {
    // Optionally specify the memory mode, indicating whether metric records can be reused or must
    // be immutable
    return MemoryMode.REUSABLE_DATA;
  }

  @Override
  public Aggregation getDefaultAggregation(InstrumentType instrumentType) {
    // Optionally specify the default aggregation as a function of instrument kind
    return Aggregation.defaultAggregation();
  }
}

MetricExporter

A MetricExporter is a plugin extension interface responsible for exporting metrics out of process. Rather than directly registering with SdkMeterProvider, they are paired with PeriodicMetricReader.

Metric exporters built-in to the SDK and maintained by the community in opentelemetry-java-contrib:

Class	Artifact	Description
`OtlpHttpMetricExporter` [1]	`io.opentelemetry:opentelemetry-exporter-otlp:1.45.0`	Exports metrics via OTLP `http/protobuf`.
`OtlpGrpcMetricExporter` [1]	`io.opentelemetry:opentelemetry-exporter-otlp:1.45.0`	Exports metrics via OTLP `grpc`.
`LoggingMetricExporter`	`io.opentelemetry:opentelemetry-exporter-logging:1.45.0`	Logs metrics to JUL in a debugging format.
`OtlpJsonLoggingMetricExporter`	`io.opentelemetry:opentelemetry-exporter-logging-otlp:1.45.0`	Logs metrics to JUL in the OTLP JSON encoding.
`OtlpStdoutMetricExporter`	`io.opentelemetry:opentelemetry-exporter-logging-otlp:1.45.0`	Logs metrics to `System.out` in the OTLP JSON file encoding (experimental).
`InterceptableMetricExporter`	`io.opentelemetry.contrib:opentelemetry-processors:1.42.0-alpha`	Passes metrics to a flexible interceptor before exporting.

[1]: See OTLP exporter sender for implementation details.

The following code snippet demonstrates MetricExporter programmatic configuration:

package otel;

import io.opentelemetry.exporter.logging.LoggingMetricExporter;
import io.opentelemetry.exporter.logging.otlp.OtlpJsonLoggingMetricExporter;
import io.opentelemetry.exporter.otlp.http.metrics.OtlpHttpMetricExporter;
import io.opentelemetry.exporter.otlp.metrics.OtlpGrpcMetricExporter;
import io.opentelemetry.sdk.metrics.export.MetricExporter;
import java.time.Duration;

public class MetricExporterConfig {
  public static MetricExporter otlpHttpMetricExporter(String endpoint) {
    return OtlpHttpMetricExporter.builder()
        .setEndpoint(endpoint)
        .addHeader("api-key", "value")
        .setTimeout(Duration.ofSeconds(10))
        .build();
  }

  public static MetricExporter otlpGrpcMetricExporter(String endpoint) {
    return OtlpGrpcMetricExporter.builder()
        .setEndpoint(endpoint)
        .addHeader("api-key", "value")
        .setTimeout(Duration.ofSeconds(10))
        .build();
  }

  public static MetricExporter logginMetricExporter() {
    return LoggingMetricExporter.create();
  }

  public static MetricExporter otlpJsonLoggingMetricExporter() {
    return OtlpJsonLoggingMetricExporter.create();
  }
}

Implement the MetricExporter interface to provide your own custom metric export logic. For example:

package otel;

import io.opentelemetry.sdk.common.CompletableResultCode;
import io.opentelemetry.sdk.common.export.MemoryMode;
import io.opentelemetry.sdk.metrics.Aggregation;
import io.opentelemetry.sdk.metrics.InstrumentType;
import io.opentelemetry.sdk.metrics.data.AggregationTemporality;
import io.opentelemetry.sdk.metrics.data.MetricData;
import io.opentelemetry.sdk.metrics.export.AggregationTemporalitySelector;
import io.opentelemetry.sdk.metrics.export.MetricExporter;
import java.util.Collection;
import java.util.logging.Level;
import java.util.logging.Logger;

public class CustomMetricExporter implements MetricExporter {

  private static final Logger logger = Logger.getLogger(CustomMetricExporter.class.getName());

  @Override
  public CompletableResultCode export(Collection<MetricData> metrics) {
    // Export the records. Typically, records are sent out of process via some network protocol, but
    // we simply log for illustrative purposes.
    logger.log(Level.INFO, "Exporting metrics");
    metrics.forEach(metric -> logger.log(Level.INFO, "Metric: " + metric));
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public CompletableResultCode flush() {
    // Export any records which have been queued up but not yet exported.
    logger.log(Level.INFO, "flushing");
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public CompletableResultCode shutdown() {
    // Shutdown the exporter and cleanup any resources.
    logger.log(Level.INFO, "shutting down");
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public AggregationTemporality getAggregationTemporality(InstrumentType instrumentType) {
    // Specify the required aggregation temporality as a function of instrument type
    return AggregationTemporalitySelector.deltaPreferred()
        .getAggregationTemporality(instrumentType);
  }

  @Override
  public MemoryMode getMemoryMode() {
    // Optionally specify the memory mode, indicating whether metric records can be reused or must
    // be immutable
    return MemoryMode.REUSABLE_DATA;
  }

  @Override
  public Aggregation getDefaultAggregation(InstrumentType instrumentType) {
    // Optionally specify the default aggregation as a function of instrument kind
    return Aggregation.defaultAggregation();
  }
}

Views

Views allow metric streams to be customized, including changing metric names, metric descriptions, metric aggregations (i.e. histogram bucket boundaries), the set of attribute keys to retain, cardinality limit, etc.

Views have somewhat unintuitive behavior when multiple match a particular instrument. If one matching view changes the metric name and another changes the metric aggregation, you might expect the name and aggregation are changed, but this is not the case. Instead, two metric streams are produced: one with the configured metric name and the default aggregation, and another with the original metric name and the configured aggregation. In other words, matching views do not merge. For best results, configure views with narrow selection criteria (i.e. select a single specific instrument).

The following code snippet demonstrates View programmatic configuration:

package otel;

import io.opentelemetry.sdk.metrics.Aggregation;
import io.opentelemetry.sdk.metrics.InstrumentSelector;
import io.opentelemetry.sdk.metrics.SdkMeterProviderBuilder;
import io.opentelemetry.sdk.metrics.View;
import java.util.List;
import java.util.Set;

public class ViewConfig {
  public static SdkMeterProviderBuilder dropMetricView(
      SdkMeterProviderBuilder builder, String metricName) {
    return builder.registerView(
        InstrumentSelector.builder().setName(metricName).build(),
        View.builder().setAggregation(Aggregation.drop()).build());
  }

  public static SdkMeterProviderBuilder histogramBucketBoundariesView(
      SdkMeterProviderBuilder builder, String metricName, List<Double> bucketBoundaries) {
    return builder.registerView(
        InstrumentSelector.builder().setName(metricName).build(),
        View.builder()
            .setAggregation(Aggregation.explicitBucketHistogram(bucketBoundaries))
            .build());
  }

  public static SdkMeterProviderBuilder attributeFilterView(
      SdkMeterProviderBuilder builder, String metricName, Set<String> keysToRetain) {
    return builder.registerView(
        InstrumentSelector.builder().setName(metricName).build(),
        View.builder().setAttributeFilter(keysToRetain).build());
  }

  public static SdkMeterProviderBuilder cardinalityLimitsView(
      SdkMeterProviderBuilder builder, String metricName, int cardinalityLimit) {
    return builder.registerView(
        InstrumentSelector.builder().setName(metricName).build(),
        View.builder().setCardinalityLimit(cardinalityLimit).build());
  }
}

SdkLoggerProvider

SdkLoggerProvider is the SDK implementation of LoggerProvider, and is responsible for handling log telemetry produced by the log bridge API.

SdkLoggerProvider is configured by the application owner, and consists of:

Resource: The resource logs are associated with.
LogRecordProcessor: Processes logs when they are emitted.
LogRecordExporter: Exports logs out of process (in conjunction with associated LogRecordProcessor).
LogLimits: Controls the limits of data associated with logs.

The following code snippet demonstrates SdkLoggerProvider programmatic configuration:

package otel;

import io.opentelemetry.sdk.logs.SdkLoggerProvider;
import io.opentelemetry.sdk.resources.Resource;

public class SdkLoggerProviderConfig {
  public static SdkLoggerProvider create(Resource resource) {
    return SdkLoggerProvider.builder()
        .setResource(resource)
        .addLogRecordProcessor(
            LogRecordProcessorConfig.batchLogRecordProcessor(
                LogRecordExporterConfig.otlpHttpLogRecordExporter("http://localhost:4318/v1/logs")))
        .setLogLimits(LogLimitsConfig::logLimits)
        .build();
  }
}

LogRecordProcessor

A LogRecordProcessor is a plugin extension interface with a callback invoked when a log is emitted. They are often paired with LogRecordExporters to export logs out of process, but have other applications such as data enrichment.

Log record processors built-in to the SDK and maintained by the community in opentelemetry-java-contrib:

Class	Artifact	Description
`BatchLogRecordProcessor`	`io.opentelemetry:opentelemetry-sdk:1.45.0`	Batches log records and exports them via a configurable `LogRecordExporter`.
`SimpleLogRecordProcessor`	`io.opentelemetry:opentelemetry-sdk:1.45.0`	Exports each log record a via a configurable `LogRecordExporter`.
`EventToSpanEventBridge`	`io.opentelemetry.contrib:opentelemetry-processors:1.42.0-alpha`	Records event log records as span events on the current span.

The following code snippet demonstrates LogRecordProcessor programmatic configuration:

package otel;

import io.opentelemetry.sdk.logs.LogRecordProcessor;
import io.opentelemetry.sdk.logs.export.BatchLogRecordProcessor;
import io.opentelemetry.sdk.logs.export.LogRecordExporter;
import io.opentelemetry.sdk.logs.export.SimpleLogRecordProcessor;
import java.time.Duration;

public class LogRecordProcessorConfig {
  public static LogRecordProcessor batchLogRecordProcessor(LogRecordExporter logRecordExporter) {
    return BatchLogRecordProcessor.builder(logRecordExporter)
        .setMaxQueueSize(2048)
        .setExporterTimeout(Duration.ofSeconds(30))
        .setScheduleDelay(Duration.ofSeconds(1))
        .build();
  }

  public static LogRecordProcessor simpleLogRecordProcessor(LogRecordExporter logRecordExporter) {
    return SimpleLogRecordProcessor.create(logRecordExporter);
  }
}

Implement the LogRecordProcessor interface to provide your own custom log processing logic. For example:

package otel;

import io.opentelemetry.api.common.AttributeKey;
import io.opentelemetry.context.Context;
import io.opentelemetry.sdk.common.CompletableResultCode;
import io.opentelemetry.sdk.logs.LogRecordProcessor;
import io.opentelemetry.sdk.logs.ReadWriteLogRecord;

public class CustomLogRecordProcessor implements LogRecordProcessor {

  @Override
  public void onEmit(Context context, ReadWriteLogRecord logRecord) {
    // Callback invoked when log record is emitted.
    // Enrich the record with a custom attribute.
    logRecord.setAttribute(AttributeKey.stringKey("my.custom.attribute"), "hello world");
  }

  @Override
  public CompletableResultCode shutdown() {
    // Optionally shutdown the processor and cleanup any resources.
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public CompletableResultCode forceFlush() {
    // Optionally process any records which have been queued up but not yet processed.
    return CompletableResultCode.ofSuccess();
  }
}

LogRecordExporter

A LogRecordExporter is a plugin extension interface responsible for exporting log records out of process. Rather than directly registering with SdkLoggerProvider, they are paired with LogRecordProcessors (typically BatchLogRecordProcessor).

Span exporters built-in to the SDK and maintained by the community in opentelemetry-java-contrib:

Class	Artifact	Description
`OtlpHttpLogRecordExporter` [1]	`io.opentelemetry:opentelemetry-exporter-otlp:1.45.0`	Exports log records via OTLP `http/protobuf`.
`OtlpGrpcLogRecordExporter` [1]	`io.opentelemetry:opentelemetry-exporter-otlp:1.45.0`	Exports log records via OTLP `grpc`.
`SystemOutLogRecordExporter`	`io.opentelemetry:opentelemetry-exporter-logging:1.45.0`	Logs log records to system out in a debugging format.
`OtlpJsonLoggingLogRecordExporter` [2]	`io.opentelemetry:opentelemetry-exporter-logging-otlp:1.45.0`	Logs log records to JUL in the OTLP JSON encoding.
`OtlpStdoutLogRecordExporter`	`io.opentelemetry:opentelemetry-exporter-logging-otlp:1.45.0`	Logs log records to `System.out` in the OTLP JSON file encoding (experimental).
`InterceptableLogRecordExporter`	`io.opentelemetry.contrib:opentelemetry-processors:1.42.0-alpha`	Passes log records to a flexible interceptor before exporting.

[1]: See OTLP exporter sender for implementation details.

[2]: OtlpJsonLoggingLogRecordExporter logs to JUL, and may cause infinite loops (i.e. JUL -> SLF4J -> Logback -> OpenTelemetry Appender -> OpenTelemetry Log SDK -> JUL) if not carefully configured.

The following code snippet demonstrates LogRecordProcessor programmatic configuration:

package otel;

import io.opentelemetry.exporter.logging.SystemOutLogRecordExporter;
import io.opentelemetry.exporter.logging.otlp.OtlpJsonLoggingLogRecordExporter;
import io.opentelemetry.exporter.otlp.http.logs.OtlpHttpLogRecordExporter;
import io.opentelemetry.exporter.otlp.logs.OtlpGrpcLogRecordExporter;
import io.opentelemetry.sdk.logs.export.LogRecordExporter;
import java.time.Duration;

public class LogRecordExporterConfig {
  public static LogRecordExporter otlpHttpLogRecordExporter(String endpoint) {
    return OtlpHttpLogRecordExporter.builder()
        .setEndpoint(endpoint)
        .addHeader("api-key", "value")
        .setTimeout(Duration.ofSeconds(10))
        .build();
  }

  public static LogRecordExporter otlpGrpcLogRecordExporter(String endpoint) {
    return OtlpGrpcLogRecordExporter.builder()
        .setEndpoint(endpoint)
        .addHeader("api-key", "value")
        .setTimeout(Duration.ofSeconds(10))
        .build();
  }

  public static LogRecordExporter systemOutLogRecordExporter() {
    return SystemOutLogRecordExporter.create();
  }

  public static LogRecordExporter otlpJsonLoggingLogRecordExporter() {
    return OtlpJsonLoggingLogRecordExporter.create();
  }
}

Implement the LogRecordExporter interface to provide your own custom log record export logic. For example:

package otel;

import io.opentelemetry.sdk.common.CompletableResultCode;
import io.opentelemetry.sdk.logs.data.LogRecordData;
import io.opentelemetry.sdk.logs.export.LogRecordExporter;
import java.util.Collection;
import java.util.logging.Level;
import java.util.logging.Logger;

public class CustomLogRecordExporter implements LogRecordExporter {

  private static final Logger logger = Logger.getLogger(CustomLogRecordExporter.class.getName());

  @Override
  public CompletableResultCode export(Collection<LogRecordData> logs) {
    // Export the records. Typically, records are sent out of process via some network protocol, but
    // we simply log for illustrative purposes.
    System.out.println("Exporting logs");
    logs.forEach(log -> System.out.println("log record: " + log));
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public CompletableResultCode flush() {
    // Export any records which have been queued up but not yet exported.
    logger.log(Level.INFO, "flushing");
    return CompletableResultCode.ofSuccess();
  }

  @Override
  public CompletableResultCode shutdown() {
    // Shutdown the exporter and cleanup any resources.
    logger.log(Level.INFO, "shutting down");
    return CompletableResultCode.ofSuccess();
  }
}

LogLimits

LogLimits defines constraints for the data captured by log records, including max attribute length, and max number of attributes.

The following code snippet demonstrates LogRecordProcessor programmatic configuration:

package otel;

import io.opentelemetry.sdk.logs.LogLimits;

public class LogLimitsConfig {
  public static LogLimits logLimits() {
    return LogLimits.builder()
        .setMaxNumberOfAttributes(128)
        .setMaxAttributeValueLength(1024)
        .build();
  }
}

TextMapPropagator

TextMapPropagator is a plugin extension interface responsible for propagating context across process boundaries in a text format.

TextMapPropagators built-in to the SDK and maintained by the community in opentelemetry-java-contrib:

Class	Artifact	Description
`W3CTraceContextPropagator`	`io.opentelemetry:opentelemetry-api:1.45.0`	Propagate trace context using W3C trace context propagation protocol.
`W3CBaggagePropagator`	`io.opentelemetry:opentelemetry-api:1.45.0`	Propagate baggage using W3C baggage propagation protocol.
`MultiTextMapPropagator`	`io.opentelemetry:opentelemetry-context:1.45.0`	Compose multiple propagators.
`JaegerPropagator`	`io.opentelemetry:opentelemetry-extension-trace-propagators:1.45.0`	Propagator trace context using the Jaeger propagation protocol.
`B3Propagator`	`io.opentelemetry:opentelemetry-extension-trace-propagators:1.45.0`	Propagator trace context using the B3 propagation protocol.
`OtTracePropagator`	`io.opentelemetry:opentelemetry-extension-trace-propagators:1.45.0`	Propagator trace context using the OpenTracing propagation protocol.
`PassThroughPropagator`	`io.opentelemetry:opentelemetry-api-incubator:1.45.0-alpha`	Propagate a configurable set fields without participating in telemetry.
`AwsXrayPropagator`	`io.opentelemetry.contrib:opentelemetry-aws-xray-propagator:1.42.0-alpha`	Propagate trace context using AWS X-Ray propagation protocol.
`AwsXrayLambdaPropagator`	`io.opentelemetry.contrib:opentelemetry-aws-xray-propagator:1.42.0-alpha`	Propagate trace context using environment variables and AWS X-Ray propagation protocol.

The following code snippet demonstrates TextMapPropagator programmatic configuration:

package otel;

import io.opentelemetry.api.baggage.propagation.W3CBaggagePropagator;
import io.opentelemetry.api.trace.propagation.W3CTraceContextPropagator;
import io.opentelemetry.context.propagation.ContextPropagators;
import io.opentelemetry.context.propagation.TextMapPropagator;

public class ContextPropagatorsConfig {
  public static ContextPropagators create() {
    return ContextPropagators.create(
        TextMapPropagator.composite(
            W3CTraceContextPropagator.getInstance(), W3CBaggagePropagator.getInstance()));
  }
}

Implement the TextMapPropagator interface to provide your own custom propagator logic. For example:

package otel;

import io.opentelemetry.context.Context;
import io.opentelemetry.context.propagation.TextMapGetter;
import io.opentelemetry.context.propagation.TextMapPropagator;
import io.opentelemetry.context.propagation.TextMapSetter;
import java.util.Collection;
import java.util.Collections;

public class CustomTextMapPropagator implements TextMapPropagator {

  @Override
  public Collection<String> fields() {
    // Return fields used for propagation. See W3CTraceContextPropagator for reference
    // implementation.
    return Collections.emptyList();
  }

  @Override
  public <C> void inject(Context context, C carrier, TextMapSetter<C> setter) {
    // Inject context. See W3CTraceContextPropagator for reference implementation.
  }

  @Override
  public <C> Context extract(Context context, C carrier, TextMapGetter<C> getter) {
    // Extract context. See W3CTraceContextPropagator for reference implementation.
    return context;
  }
}

Appendix

Internal logging

SDK components log a variety of information to java.util.logging, at different log levels and using logger names based on the fully qualified class name of the relevant component.

By default, log messages are handled by the root handler in your application. If you have not installed a custom root handler for your application, logs of level INFO or higher are sent to the console by default.

You may want to change the behavior of the logger for OpenTelemetry. For example, you can reduce the logging level to output additional information when debugging, increase the level for a particular class to ignore errors coming from the class, or install a custom handler or filter to run custom code whenever OpenTelemetry logs a particular message. No detailed list of logger names and log information is maintained. However, all OpenTelemetry API, SDK, contrib and instrumentation components share the same io.opentelemetry.* package prefix. It can be useful to enable finer grain logs for all io.opentelemetry.*, inspect the output, and narrow down to packages or FQCNs of interest.

For example:

## Turn off all OpenTelemetry logging
io.opentelemetry.level = OFF

## Turn off logging for just the BatchSpanProcessor
io.opentelemetry.sdk.trace.export.BatchSpanProcessor.level = OFF

## Log "FINE" messages for help in debugging
io.opentelemetry.level = FINE

## Sets the default ConsoleHandler's logger's level
## Note this impacts the logging outside of OpenTelemetry as well
java.util.logging.ConsoleHandler.level = FINE

For more fine-grained control and special case handling, custom handlers and filters can be specified with code.

// Custom filter which does not log errors which come from the export
public class IgnoreExportErrorsFilter implements java.util.logging.Filter {

 public boolean isLoggable(LogRecord record) {
    return !record.getMessage().contains("Exception thrown by the export");
 }
}

## Registering the custom filter on the BatchSpanProcessor
io.opentelemetry.sdk.trace.export.BatchSpanProcessor = io.opentelemetry.extension.logging.IgnoreExportErrorsFilter

OTLP exporter senders

The span exporter, metric exporter, and log exporter discuss OTLP exporters of the form:

OtlpHttp{Signal}Exporters export data via OTLP http/protobuf.
OtlpGrpc{Signal}Exporters export data via OTLP grpc.

The exporters for all signals are available via io.opentelemetry:opentelemetry-exporter-otlp:1.45.0.

Internally, these exporters depend on various client libraries to execute HTTP and gRPC requests. There is no single HTTP / gRPC client library which satisfies all use cases in the Java ecosystem:

Java 11+ brings the built-in java.net.http.HttpClient, but opentelemetry-java needs to support Java 8+ users, and this can’t be used to export via gRPC because there is no support for trailer headers.
OkHttp provides a powerful HTTP client with support for trailer headers, but depends on the kotlin standard library.
grpc-java provides its own ManagedChannel abstraction with various transport implementations, but is not suitable for http/protobuf.

In order to accommodate various use cases, opentelemetry-exporter-otlp uses an internal “sender” abstraction, with a variety of implementations to reflect application constraints. To choose another implementation, exclude the io.opentelemetry:opentelemetry-exporter-sender-okhttp default dependency, and add a dependency on the alternative.

Artifact	Description	OTLP Protocols	Default
`io.opentelemetry:opentelemetry-exporter-sender-okhttp:1.45.0`	OkHttp based implementation.	`grpc`, `http/protobuf`	Yes
`io.opentelemetry:opentelemetry-exporter-sender-jdk:1.45.0`	Java 11+ `java.net.http.HttpClient` based implementation.	`http/protobuf`	No
`io.opentelemetry:opentelemetry-exporter-sender-grpc-managed-channel:1.45.0` [1]	`grpc-java` `ManagedChannel` based implementation.	`grpc`	No

[1]: In order to use opentelemetry-exporter-sender-grpc-managed-channel, you must also add a dependency on a gRPC transport implementations.

Testing

TODO: document tools available for testing the SDK

[i18n] feedback_title

[i18n] feedback_question

Thank you. Your feedback is appreciated!

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

最終更新 November 25, 2024: Add missing contrib processors, add auto version update for opentelemetry-java-contrib (#5664) (861c566e)