NutriFlow — Trazabilidad Distribuida con Jaeger

Trazas / hora

1,247

↑ 12% vs. última hora

Latencia P95 global

2.31s

POST /api/plans/generate

Trazas con error

3.4%

42 trazas en la última hora

Servicios instrumentados

4 / 4

OpenTelemetry SDK activo

TRAZA REPRESENTATIVA · POST /api/plans/generate

Trace ID: 4bf92f3577b34da6a3ce929d0e0e4736

Duración total 2.847s

Spans 14

Servicios 4

Estado 1 error

Servicio · Operación

Timeline (0 → 2847ms)

api-gateway · HTTP POST /api/plans/generate

2847ms

auth_middleware (JWT validate)

80ms

rate_limit_check (Redis)

22ms

patient-service · GET /internal/patients/:id

267ms

db.query — SELECT patients (PostgreSQL)

127ms

db.query — SELECT nutrition_history

99ms

nutrition-service · POST /internal/plans/generate

2332ms ← cuello de botella

validate_plan_input

19ms

cache.get — Redis (MISS)

14ms

ai_model_inference · NutriGPT v2.1 (CPU)

2165ms ⚠ P99: 5.8s

db.query — INSERT nutrition_plans

71ms

notification-service · POST /internal/notify (email)

160ms

serialize_response (JSON)

28ms

0 284ms 569ms 853ms 1.14s 1.42s 1.71s 1.99s 2.28s 2.56s 2.85s

Servicios Instrumentados

🔀

api-gateway

Node.js 20 Express 4.x JWT OTLP gRPC

108ms

P99 · 1,247 req/h

🥗

nutrition-service

Python 3.11 Flask 3.x SQLAlchemy 100% sampling

5.8s

P99 · 1,247 req/h

👤

patient-service

Python 3.11 Flask 3.x PostgreSQL 15

312ms

P99 · 1,247 req/h

🔔

notification-service

Node.js 20 SMTP Push API

198ms

P99 · 1,247 req/h

Estrategia de Muestreo (jaeger-sampling.json)

nutrition-service

probabilistic · ruta crítica IA

100%

api-gateway

probabilistic

50%

patient-service

probabilistic (default)

10%

notification-service

probabilistic (default)

10%

GET /health, /ready

operation override

0.1%

Distribución de latencia — última hora

<100ms

187

100-500

312

500-1s

431

1-3s

218

3-6s

>6s

INSTRUMENTACION · nutrition-service/tracing.py

OpenTelemetry SDK · Python 3.11 · Exporter: OTLP gRPC → Jaeger :4317

Python

# nutrition-service/tracing.py — NutriFlow
from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter
from opentelemetry.sdk.resources import Resource
from opentelemetry.instrumentation.flask import FlaskInstrumentor
from opentelemetry.instrumentation.sqlalchemy import SQLAlchemyInstrumentor
from opentelemetry.instrumentation.redis import RedisInstrumentor

def init_tracing(service_name: str):
    resource = Resource.create({
        "service.name": service_name,
        "service.version": "2.1.0",
        "deployment.environment": "staging",
        "team": "nutriflow-backend",
    })
    provider = TracerProvider(resource=resource)
    exporter = OTLPSpanExporter(endpoint="http://jaeger:4317", insecure=True)
    provider.add_span_processor(BatchSpanProcessor(exporter))
    trace.set_tracer_provider(provider)

    FlaskInstrumentor().instrument()
    SQLAlchemyInstrumentor().instrument()
    RedisInstrumentor().instrument()
    return trace.get_tracer(service_name)

# app.py — span personalizado para la inferencia IA
tracer = init_tracing("nutrition-service")

@app.route("/internal/plans/generate", methods=["POST"])
def generate_plan():
    with tracer.start_as_current_span("ai_model_inference") as span:
        span.set_attribute("model.name", "NutriGPT-v2.1")
        span.set_attribute("model.device", "cpu")          # ← candidato a GPU
        span.set_attribute("patient.goal", data["goal"])
        span.set_attribute("plan.days", 7)
        result = nutrigpt_model.generate(data)          # P99: 5.8s en CPU
        span.set_attribute("plan.tokens_generated", result["tokens"])
    return result, 200

NutriFlow — Trazabilidad Distribuida

Cuello de botella detectado — nutrition-service · span: ai_model_inference