分布式链路追踪：Jaeger与Zipkin架构

分布式追踪原理

分布式链路追踪通过在请求入口生成唯一Trace ID，贯穿整个调用链，记录每个Span的耗时和状态，用于分析分布式系统的性能瓶颈。

用户请求 → API Gateway → Service A → Service B → Database
    │          │            │           │           │
    └──────────┴────────────┴───────────┴───────────┘
                    Trace ID: abc123
                    
Span树结构：
Trace (abc123)
├── Span: API Gateway (100ms)
│   └── Span: Service A (80ms)
│       ├── Span: Service B (50ms)
│       │   └── Span: Database Query (20ms)
│       └── Span: Cache Lookup (5ms)

Jaeger架构

核心组件

Jaeger架构：
┌─────────────────────────────────────────────────┐
│              Jaeger Architecture                 │
├─────────────────┬───────────────────────────────┤
│  Agent          │  Collector                    │
│  接收Span       │  处理、验证、索引              │
│  UDP接收        │  写入存储                     │
├─────────────────┴───────────────────────────────┤
│              Query Service                      │
│         查询API + Web UI                         │
├─────────────────────────────────────────────────┤
│              Storage Backend                     │
│        Elasticsearch / Cassandra / Kafka        │
└─────────────────────────────────────────────────┘

Jaeger部署（Kubernetes）

# jaeger-deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: jaeger
spec:
  replicas: 1
  selector:
    matchLabels:
      app: jaeger
  template:
    metadata:
      labels:
        app: jaeger
    spec:
      containers:
      - name: jaeger
        image: jaegertracing/all-in-one:latest
        ports:
        - containerPort: 16686  # UI
        - containerPort: 14268  # HTTP collector
        - containerPort: 6831   # Agent UDP
        env:
        - name: COLLECTOR_STORAGE_TYPE
          value: elasticsearch
        - name: ES_SERVER_URLS
          value: http://elasticsearch:9200
---
apiVersion: v1
kind: Service
metadata:
  name: jaeger
spec:
  selector:
    app: jaeger
  ports:
  - name: ui
    port: 16686
  - name: collector
    port: 14268

Zipkin架构

核心组件

Zipkin架构：
┌─────────────────────────────────────────────────┐
│                Zipkin                            │
├─────────────────┬───────────────────────────────┤
│  Collector      │  Storage                       │
│  收集Span数据   │  存储追踪数据                  │
├─────────────────┴───────────────────────────────┤
│  API + UI                                        │
│  查询和可视化                                    │
└─────────────────────────────────────────────────┘

Zipkin配置

# docker-compose.yml
version: '3'
services:
  zipkin:
    image: openzipkin/zipkin:latest
    ports:
      - "9411:9411"
    environment:
      - STORAGE_TYPE=elasticsearch
      - ES_HOSTS=http://elasticsearch:9200
    depends_on:
      - elasticsearch
  
  elasticsearch:
    image: docker.elastic.co/elasticsearch/elasticsearch:8.11.0
    environment:
      - discovery.type=single-node
      - xpack.security.enabled=false
    ports:
      - "9200:9200"

Trace传播机制

W3C Trace Context

# HTTP Header传播
# 请求头格式：
# traceparent: 00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01
# tracestate: vendor1=value1,vendor2=value2

import requests
from opentelemetry.propagate import inject, extract

def call_downstream_service():
    headers = {}
    inject(headers)  # 注入当前context到headers
    
    response = requests.get(
        "http://downstream-service/api",
        headers=headers
    )
    return response

# 接收端提取context
def handle_request(request):
    context = extract(request.headers)  # 从headers提取context
    with tracer.start_as_current_span("handle_request", context=context):
        process_request()

Baggage传播

from opentelemetry import baggage, context

# 设置Baggage（跨服务传播的键值对）
ctx = baggage.set_baggage("user.id", "user123")
ctx = baggage.set_baggage("request.id", "req456", context=ctx)

# 传播Baggage
headers = {}
inject(headers, context=ctx)

# 下游服务读取Baggage
ctx = extract(request.headers)
user_id = baggage.get_baggage("user.id", context=ctx)

采样策略

# 采样配置
sampling_strategies:
  # 概率采样
  probabilistic:
    type: probabilistic
    param: 0.1  # 10%采样率
  
  # 限速采样
  rateLimiting:
    type: rateLimiting
    param: 100  # 每秒100条
  
  # 尾部采样（基于延迟/错误）
  tail_sampling:
    policies:
      - name: error-based
        type: status_code
        status_code:
          status_codes: [ERROR]
      - name: latency-based
        type: latency
        latency:
          threshold_ms: 1000

最佳实践

1. 采样策略：高QPS服务使用概率采样，关键路径使用尾部采样
2. 上下文传播：确保所有HTTP/gRPC调用正确传播Trace Context
3. Span命名：使用有意义的Span名称，如GET /api/users/{id}
4. 错误记录：在Span中记录错误堆栈，便于问题定位