codewithbehnam/distributed-tracing

Distributed Tracing

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

Purpose

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

When to Use

• 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 - <<EOF
apiVersion: jaegertracing.io/v1
kind: Jaeger
metadata:
  name: jaeger
  namespace: observability
spec:
  strategy: production
  storage:
    type: elasticsearch
    options:
      es:
        server-urls: http://elasticsearch:9200
  ingress:
    enabled: true
EOF

Docker Compose

version: "3.8"
services:
  jaeger:
    image: jaegertracing/all-in-one:latest
    ports:
      - "5775:5775/udp"
      - "6831:6831/udp"
      - "6832:6832/udp"
      - "5778:5778"
      - "16686:16686" # UI
      - "14268:14268" # Collector
      - "14250:14250" # gRPC
      - "9411:9411" # Zipkin
    environment:
      - COLLECTOR_ZIPKIN_HOST_PORT=:9411

Reference: See references/jaeger-setup.md

Application Instrumentation

OpenTelemetry (Recommended)

Python (Flask)

from opentelemetry import trace
from opentelemetry.exporter.jaeger.thrift import JaegerExporter
from opentelemetry.sdk.resources import SERVICE_NAME, Resource
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from opentelemetry.instrumentation.flask import FlaskInstrumentor
from flask import Flask

# Initialize tracer
resource = Resource(attributes={SERVICE_NAME: "my-service"})
provider = TracerProvider(resource=resource)
processor = BatchSpanProcessor(JaegerExporter(
    agent_host_name="jaeger",
    agent_port=6831,
))
provider.add_span_processor(processor)
trace.set_tracer_provider(provider)

# Instrument Flask
app = Flask(__name__)
FlaskInstrumentor().instrument_app(app)

@app.route('/api/users')
def get_users():
    tracer = trace.get_tracer(__name__)

    with tracer.start_as_current_span("get_users") as span:
        span.set_attribute("user.count", 100)
        # Business logic
        users = fetch_users_from_db()
        return {"users": users}

def fetch_users_from_db():
    tracer = trace.get_tracer(__name__)

    with tracer.start_as_current_span("database_query") as span:
        span.set_attribute("db.system", "postgresql")
        span.set_attribute("db.statement", "SELECT * FROM users")
        # Database query
        return query_database()

Node.js (Express)

const { NodeTracerProvider } = require("@opentelemetry/sdk-trace-node");
const { JaegerExporter } = require("@opentelemetry/exporter-jaeger");
const { BatchSpanProcessor } = require("@opentelemetry/sdk-trace-base");
const { registerInstrumentations } = require("@opentelemetry/instrumentation");
const { HttpInstrumentation } = require("@opentelemetry/instrumentation-http");
const {
  ExpressInstrumentation,
} = require("@opentelemetry/instrumentation-express");

// Initialize tracer
const provider = new NodeTracerProvider({
  resource: { attributes: { "service.name": "my-service" } },
});

const exporter = new JaegerExporter({
  endpoint: "http://jaeger:14268/api/traces",
});

provider.addSpanProcessor(new BatchSpanProcessor(exporter));
provider.register();

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

const express = require("express");
const app = express();

app.get("/api/users", async (req, res) => {
  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

Related Skills

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