Distributed Tracing

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 - <<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

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

---

<!-- AGI-INTEGRATION-START -->

AGI Framework Integration

Adapted for @techwavedev/​agi-agent-kit Original source: antigravity-awesome-skills

Memory-First Protocol

Retrieve prior error resolutions and debugging strategies. The hybrid search excels here — BM25 finds exact error codes/​stack traces while vectors find semantically similar past issues.

# Check for prior debugging/​diagnostics context before starting
python3 execution/​memory_manager.py auto --query "error patterns and debugging solutions for Distributed Tracing"

Storing Results

After completing work, store debugging/​diagnostics decisions for future sessions:

python3 execution/​memory_manager.py store \
  --content "Root cause: memory leak from unclosed DB connections in pool — fixed with context manager" \
  --type error --project <project> \
  --tags distributed-tracing debugging

Multi-Agent Collaboration

Store error resolutions so any agent encountering the same issue retrieves the fix instantly instead of re-debugging.

python3 execution/​cross_agent_context.py store \
  --agent "<your-agent>" \
  --action "Debugged and resolved critical issue — root cause documented for future reference" \
  --project <project>

Self-Annealing Loop

When this skill resolves an error, store the fix in memory AND update the relevant directive. The system gets stronger with each resolved issue.

BM25 Exact Match

Error codes, stack traces, and log messages are best found via BM25 keyword search. The hybrid system automatically uses exact matching for these patterns.

<!-- AGI-INTEGRATION-END -->