OpenTelemetry with Spring Boot: Modern Distributed Tracing

Eleftheria DrosopoulouApril 11th, 2025Last Updated: April 12th, 2025

0 168 2 minutes read

In modern microservices architectures, tracking requests as they flow across service boundaries is critical for debugging performance bottlenecks, diagnosing failures, and maintaining system reliability. While Zipkin and Jaeger have been popular distributed tracing tools, OpenTelemetry has emerged as the new standard, offering vendor-neutral instrumentation, logs, metrics, and traces in a unified framework.

This article explores:

Why OpenTelemetry (OTel) is replacing traditional tracing tools
How to integrate OpenTelemetry with Spring Boot
Best practices for distributed tracing in microservices
Visualizing traces with Jaeger, Zipkin, and Grafana

1. Why OpenTelemetry? The Evolution Beyond Zipkin

The Limitations of Traditional Tracing (Zipkin/Jaeger)

Vendor lock-in: Zipkin/Jaeger require specific SDKs.
Limited correlation: Metrics, logs, and traces were separate.
Manual instrumentation: Required explicit code changes.

How OpenTelemetry Solves These Problems

Standardized APIs: Works across multiple backends (Jaeger, Zipkin, Prometheus, etc.).
Auto-instrumentation: Captures traces without code changes.
Unified Observability: Combines traces, metrics, and logs.

2. Setting Up OpenTelemetry in Spring Boot

Step 1: Add Dependencies

<!-- OpenTelemetry SDK & Auto-Instrumentation -->
<dependency>
    <groupId>io.opentelemetry</groupId>
    <artifactId>opentelemetry-api</artifactId>
    <version>1.30.0</version>
</dependency>
<dependency>
    <groupId>io.opentelemetry</groupId>
    <artifactId>opentelemetry-sdk</artifactId>
    <version>1.30.0</version>
</dependency>
<dependency>
    <groupId>io.opentelemetry.instrumentation</groupId>
    <artifactId>opentelemetry-spring-boot-starter</artifactId>
    <version>2.1.0</version>
</dependency>

Step 2: Configure application.yml

opentelemetry:
  service:
    name: order-service
  traces:
    exporter:
      otlp:
        endpoint: http://otel-collector:4317 # OTLP Collector
  logs:
    exporter:
      otlp:
        enabled: true
  metrics:
    exporter:
      otlp:
        enabled: true

Step 3: Deploy an OpenTelemetry Collector

# docker-compose.yml
services:
  otel-collector:
    image: otel/opentelemetry-collector
    ports:
      - "4317:4317"   # OTLP gRPC
      - "4318:4318"   # OTLP HTTP
    volumes:
      - ./otel-config.yaml:/etc/otel-config.yaml
    command: ["--config=/etc/otel-config.yaml"]

Step 4: Export Traces to Jaeger/Zipkin

# otel-config.yaml
exporters:
  jaeger:
    endpoint: "jaeger:14250"
    tls:
      insecure: true
  zipkin:
    endpoint: "http://zipkin:9411/api/v2/spans"
 
service:
  pipelines:
    traces:
      receivers: [otlp]
      processors: [batch]
      exporters: [jaeger, zipkin]

3. Advanced Distributed Tracing Techniques

Custom Spans for Business Logic

@GetMapping("/process")
public String processOrder() {
    Span span = Span.current();
    span.setAttribute("order.id", "12345"); // Custom attributes
     
    try (Scope scope = span.makeCurrent()) {
        // Business logic
        inventoryService.reserveItems();
        paymentService.chargeCustomer();
        return "Order processed!";
    } catch (Exception e) {
        span.recordException(e); // Error tracking
        throw e;
    }
}

Baggage Propagation (Cross-Service Context)

// Set baggage in Service A
Baggage.current()
    .toBuilder()
    .put("user.id", "user-123")
    .build()
    .makeCurrent();
 
// Retrieve in Service B
String userId = Baggage.current().getEntryValue("user.id");

Sampling Strategies

# application.yml
opentelemetry:
  traces:
    sampler: parentbased_always_on # or "dynamic" for adaptive sampling

4. Visualizing Traces: Jaeger vs. Zipkin vs. Grafana

Tool	Strengths	Best For
Jaeger	Powerful querying, dependency graphs	Debugging complex microservices
Zipkin	Simple UI, lightweight	Basic tracing needs
Grafana Tempo	Deep integration with Prometheus metrics	Full observability (logs + metrics + traces)

5. Best Practices for Production

Use OTLP (OpenTelemetry Protocol) instead of vendor-specific exporters.
Enable Auto-Instrumentation for Spring Web, JDBC, Kafka, etc.
Implement Sampling to reduce storage costs.
Correlate Logs & Traces using trace_id.
Monitor Collector Performance to avoid bottlenecks.

Conclusion

OpenTelemetry provides a future-proof, vendor-neutral way to implement distributed tracing in Spring Boot microservices. By replacing Zipkin/Jaeger with OTel, teams gain: