Distributed Tracing

Distributed Tracing

Comprehensive distributed tracing with Jaeger, Zipkin, OpenTelemetry, correlation IDs, and span design.

Overview

Distributed tracing tracks requests as they flow through multiple services, enabling debugging and performance analysis in microservices architectures.

Key Concepts

Trace Model

• Trace: End-to-end request journey
• Span: Single operation within a trace
• Span Context: Propagated trace information
• Baggage: Custom key-value pairs carried across services

Span Attributes

• Operation Name: What the span represents
• Start/End Time: Duration measurement
• Tags: Indexed metadata for querying
• Logs: Time-stamped events within span
• Status: Success, error, or unset

OpenTelemetry Implementation

Instrumentation Setup

// Node.js OpenTelemetry setup
const { NodeTracerProvider } = require('@opentelemetry/sdk-trace-node');
const { SimpleSpanProcessor } = require('@opentelemetry/sdk-trace-base');
const { JaegerExporter } = require('@opentelemetry/exporter-jaeger');
const { registerInstrumentations } = require('@opentelemetry/instrumentation');
const { HttpInstrumentation } = require('@opentelemetry/instrumentation-http');
const { ExpressInstrumentation } = require('@opentelemetry/instrumentation-express');

const provider = new NodeTracerProvider();

provider.addSpanProcessor(
  new SimpleSpanProcessor(
    new JaegerExporter({
      endpoint: 'http://jaeger:14268/api/traces',
    })
  )
);

provider.register();

registerInstrumentations({
  instrumentations: [
    new HttpInstrumentation(),
    new ExpressInstrumentation(),
  ],
});

Manual Span Creation

const { trace } = require('@opentelemetry/api');

const tracer = trace.getTracer('my-service');

async function processOrder(orderId) {
  return tracer.startActiveSpan('processOrder', async (span) => {
    try {
      span.setAttribute('order.id', orderId);

      // Child span for database operation
      await tracer.startActiveSpan('db.query', async (dbSpan) => {
        dbSpan.setAttribute('db.system', 'postgresql');
        dbSpan.setAttribute('db.statement', 'SELECT * FROM orders WHERE id = $1');
        await db.query('SELECT * FROM orders WHERE id = $1', [orderId]);
        dbSpan.end();
      });

      span.setStatus({ code: SpanStatusCode.OK });
    } catch (error) {
      span.setStatus({ code: SpanStatusCode.ERROR, message: error.message });
      span.recordException(error);
      throw error;
    } finally {
      span.end();
    }
  });
}

Context Propagation

const { context, propagation } = require('@opentelemetry/api');

// Extract context from incoming request
app.use((req, res, next) => {
  const ctx = propagation.extract(context.active(), req.headers);
  context.with(ctx, next);
});

// Inject context into outgoing request
async function callService(url) {
  const headers = {};
  propagation.inject(context.active(), headers);

  return fetch(url, { headers });
}

Jaeger Configuration

Kubernetes Deployment

apiVersion: jaegertracing.io/v1
kind: Jaeger
metadata:
  name: jaeger
spec:
  strategy: production
  storage:
    type: elasticsearch
    elasticsearch:
      nodeCount: 3
      resources:
        requests:
          cpu: 1
          memory: 4Gi
  collector:
    maxReplicas: 5
  query:
    replicas: 2

Sampling Strategies

# Jaeger sampling configuration
apiVersion: v1
kind: ConfigMap
metadata:
  name: jaeger-sampling
data:
  sampling: |
    {
      "service_strategies": [
        {
          "service": "order-service",
          "type": "probabilistic",
          "param": 0.5
        },
        {
          "service": "payment-service",
          "type": "ratelimiting",
          "param": 100
        }
      ],
      "default_strategy": {
        "type": "probabilistic",
        "param": 0.1
      }
    }

Span Design Guidelines

Naming Conventions

HTTP spans:    HTTP {METHOD} {route}
               HTTP GET /api/users/:id

Database:      {db.system}.{operation}
               postgresql.query

Message:       {messaging.system} {operation} {destination}
               kafka send orders-topic

RPC:           {rpc.system}/{service}/{method}
               grpc/UserService/GetUser

Essential Attributes

// HTTP spans
span.setAttribute('http.method', 'GET');
span.setAttribute('http.url', 'https://api.example.com/users/123');
span.setAttribute('http.status_code', 200);
span.setAttribute('http.request_content_length', 0);
span.setAttribute('http.response_content_length', 1234);

// Database spans
span.setAttribute('db.system', 'postgresql');
span.setAttribute('db.name', 'mydb');
span.setAttribute('db.statement', 'SELECT * FROM users WHERE id = $1');
span.setAttribute('db.operation', 'SELECT');

// Messaging spans
span.setAttribute('messaging.system', 'kafka');
span.setAttribute('messaging.destination', 'orders');
span.setAttribute('messaging.operation', 'send');

Best Practices

1. Consistent Naming: Follow semantic conventions
2. Don't Over-Trace: Sample appropriately
3. Meaningful Spans: Business-relevant operations
4. Error Recording: Always record exceptions
5. Context Propagation: Ensure trace continuity

Sampling Strategies

Head-Based Sampling

• Decision made at trace start
• Simpler, consistent
• May miss interesting traces

Tail-Based Sampling

• Decision made at trace end
• Keeps all errors and slow traces
• More resource intensive

Adaptive Sampling

• Adjusts rate based on traffic
• Balances cost and coverage
• Best for variable traffic

Anti-Patterns

• Creating spans for every function call
• Not propagating context across service boundaries
• Ignoring span errors
• Sampling 100% in production
• Not correlating traces with logs

When to Use

• Microservices with complex request flows
• Debugging latency issues
• Understanding service dependencies
• Capacity planning

When NOT to Use

• Monolithic applications
• Very high-throughput systems without sampling
• When storage costs are a concern

Converted and distributed by TomeVault — claim your Tome and manage your conversions.