Jaeger — Visualizing Distributed Traces

The Problem

You have traces being generated. Now you need to see them. Jaeger is an open-source distributed tracing backend — it stores traces and provides a UI for searching and visualizing them.

Running Jaeger

# infrastructure/docker-compose.yml (relevant service)
services:
  jaeger:
    image: jaegertracing/all-in-one:1.53
    container_name: jaeger
    environment:
      COLLECTOR_OTLP_ENABLED: "true"
    ports:
      - "16686:16686"  # Jaeger UI
      - "4317:4317"    # OTLP gRPC receiver
      - "4318:4318"    # OTLP HTTP receiver

The all-in-one image includes collector, query service, and UI — perfect for development and small-scale production.

Connecting Your App

OTEL_ENABLED=true
OTEL_EXPORTER_OTLP_ENDPOINT=http://jaeger:4317

Your app sends traces via OTLP gRPC to port 4317. Jaeger stores them in memory (development) or a backing store like Elasticsearch or Cassandra (production).

Using the Jaeger UI

Open http://localhost:16686:

1. Service dropdown — Select python-production-blueprint
2. Operation dropdown — Filter by endpoint (e.g., POST /api/v1/items/)
3. Search — Find traces by time range, duration, tags
4. Trace detail — Click a trace for the waterfall diagram

Reading the Waterfall

POST /api/v1/items/  ████████████████████████  45ms
  validate_input     ██                         2ms
  DB INSERT          ██████████████████         30ms
  log_event          █                          1ms

The DB insert is the bottleneck — 67% of request time. Without tracing, you’d be guessing.

Searching by Tags

Jaeger for Multiple Services

If you have Service A calling Service B:

Service A: POST /api/v1/orders/
├── Span: validate_order (5ms)
├── Span: HTTP POST service-b/api/v1/inventory/ (120ms)  ← crosses service boundary
│   └── Service B: POST /api/v1/inventory/
│       ├── Span: check_stock (15ms)
│       └── Span: DB SELECT inventory (95ms)
└── Span: DB INSERT orders (25ms)

The traceparent header carries the trace context. Service B’s spans appear as children of Service A’s HTTP call span.

Production Jaeger

For production, use a distributed backend:

jaeger:
  image: jaegertracing/all-in-one:1.53
  environment:
    SPAN_STORAGE_TYPE: elasticsearch
    ES_SERVER_URLS: http://elasticsearch:9200

Or use Grafana Tempo as the backend — it’s designed for high-volume trace storage.

Next Step

In the next lesson, we add Prometheus metrics — the RED method for monitoring Rate, Errors, and Duration.