Distributed Tracing

Implement distributed tracing with Jaeger and Tempo for request flow visibility across microservices.

Do not use this skill when

• The task is unrelated to distributed tracing
• You need a different domain or tool outside this scope

Instructions

• Clarify goals, constraints, and required inputs.
• Apply relevant best practices and validate outcomes.
• Provide actionable steps and verification.
• If detailed examples are required, open resources/implementation-playbook.md.

Purpose

Track requests across distributed systems to understand latency, dependencies, and failure points.

Use this skill when

• Debug latency issues
• Understand service dependencies
• Identify bottlenecks
• Trace error propagation
• Analyze request paths

Distributed Tracing Concepts

Trace Structure

Trace (Request ID: abc123)
  ↓
Span (frontend) [100ms]
  ↓
Span (api-gateway) [80ms]
  ├→ Span (auth-service) [10ms]
  └→ Span (user-service) [60ms]
      └→ Span (database) [40ms]

Key Components

• Trace - End-to-end request journey
• Span - Single operation within a trace
• Context - Metadata propagated between services
• Tags - Key-value pairs for filtering
• Logs - Timestamped events within a span

Jaeger Setup

Kubernetes Deployment

# Deploy Jaeger Operator
kubectl create namespace observability
kubectl create -f https://github.com/jaegertracing/jaeger-operator/releases/download/v1.51.0/jaeger-operator.yaml -n observability

# Deploy Jaeger instance
kubectl apply -f -  {
  const tracer = trace.getTracer('my-service');
  const span = tracer.startSpan('get_users');

  try {
    const users = await fetchUsers();
    span.setAttributes({ 'user.count': users.length });
    res.json({ users });
  } finally {
    span.end();
  }
});

Go

package main

import (
    "context"
    "go.opentelemetry.io/otel"
    "go.opentelemetry.io/otel/exporters/jaeger"
    "go.opentelemetry.io/otel/sdk/resource"
    sdktrace "go.opentelemetry.io/otel/sdk/trace"
    semconv "go.opentelemetry.io/otel/semconv/v1.4.0"
)

func initTracer() (*sdktrace.TracerProvider, error) {
    exporter, err := jaeger.New(jaeger.WithCollectorEndpoint(
        jaeger.WithEndpoint("http://jaeger:14268/api/traces"),
    ))
    if err != nil {
        return nil, err
    }

    tp := sdktrace.NewTracerProvider(
        sdktrace.WithBatcher(exporter),
        sdktrace.WithResource(resource.NewWithAttributes(
            semconv.SchemaURL,
            semconv.ServiceNameKey.String("my-service"),
        )),
    )

    otel.SetTracerProvider(tp)
    return tp, nil
}

func getUsers(ctx context.Context) ([]User, error) {
    tracer := otel.Tracer("my-service")
    ctx, span := tracer.Start(ctx, "get_users")
    defer span.End()

    span.SetAttributes(attribute.String("user.filter", "active"))

    users, err := fetchUsersFromDB(ctx)
    if err != nil {
        span.RecordError(err)
        return nil, err
    }

    span.SetAttributes(attribute.Int("user.count", len(users)))
    return users, nil
}

Reference: See references/instrumentation.md

Context Propagation

HTTP Headers

traceparent: 00-0af7651916cd43dd8448eb211c80319c-b7ad6b7169203331-01
tracestate: congo=t61rcWkgMzE

Propagation in HTTP Requests

Python

from opentelemetry.propagate import inject

headers = {}
inject(headers)  # Injects trace context

response = requests.get('http://downstream-service/api', headers=headers)

Node.js

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

const headers = {};
propagation.inject(context.active(), headers);

axios.get('http://downstream-service/api', { headers });

Tempo Setup (Grafana)

Kubernetes Deployment

apiVersion: v1
kind: ConfigMap
metadata:
  name: tempo-config
data:
  tempo.yaml: |
    server:
      http_listen_port: 3200

    distributor:
      receivers:
        jaeger:
          protocols:
            thrift_http:
            grpc:
        otlp:
          protocols:
            http:
            grpc:

    storage:
      trace:
        backend: s3
        s3:
          bucket: tempo-traces
          endpoint: s3.amazonaws.com

    querier:
      frontend_worker:
        frontend_address: tempo-query-frontend:9095
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: tempo
spec:
  replicas: 1
  template:
    spec:
      containers:
      - name: tempo
        image: grafana/tempo:latest
        args:
          - -config.file=/etc/tempo/tempo.yaml
        volumeMounts:
        - name: config
          mountPath: /etc/tempo
      volumes:
      - name: config
        configMap:
          name: tempo-config

Reference: See assets/jaeger-config.yaml.template

Sampling Strategies

Probabilistic Sampling

# Sample 1% of traces
sampler:
  type: probabilistic
  param: 0.01

Rate Limiting Sampling

# Sample max 100 traces per second
sampler:
  type: ratelimiting
  param: 100

Adaptive Sampling

from opentelemetry.sdk.trace.sampling import ParentBased, TraceIdRatioBased

# Sample based on trace ID (deterministic)
sampler = ParentBased(root=TraceIdRatioBased(0.01))

Trace Analysis

Finding Slow Requests

Jaeger Query:

service=my-service
duration > 1s

Finding Errors

Jaeger Query:

service=my-service
error=true
tags.http.status_code >= 500

Service Dependency Graph

Jaeger automatically generates service dependency graphs showing:

• Service relationships
• Request rates
• Error rates
• Average latencies

Best Practices

1. Sample appropriately (1-10% in production)
2. Add meaningful tags (user_id, request_id)
3. Propagate context across all service boundaries
4. Log exceptions in spans
5. Use consistent naming for operations
6. Monitor tracing overhead (<1% CPU impact)
7. Set up alerts for trace errors
8. Implement distributed context (baggage)
9. Use span events for important milestones
10. Document instrumentation standards

Integration with Logging

Correlated Logs

import logging
from opentelemetry import trace

logger = logging.getLogger(__name__)

def process_request():
    span = trace.get_current_span()
    trace_id = span.get_span_context().trace_id

    logger.info(
        "Processing request",
        extra={"trace_id": format(trace_id, '032x')}
    )

Troubleshooting

No traces appearing:

• Check collector endpoint
• Verify network connectivity
• Check sampling configuration
• Review application logs

High latency overhead:

• Reduce sampling rate
• Use batch span processor
• Check exporter configuration

Reference Files

• references/jaeger-setup.md - Jaeger installation
• references/instrumentation.md - Instrumentation patterns
• assets/jaeger-config.yaml.template - Jaeger configuration

Related Skills

• prometheus-configuration - For metrics
• grafana-dashboards - For visualization
• slo-implementation - For latency SLOs