API网关设计：从单点到高可用的架构演进-育师

前言

API网关是微服务架构中的关键组件。我们从一个简单的Nginx反向代理，演进到一个功能完整的API网关系统。这个过程中，我们学到了很多。

一、问题的开始

最初，我们用Nginx做反向代理：

nginx

upstream backend { server app1:8080; server app2:8080; server app3:8080; } server { listen 80; location / { proxy_pass http://backend; } }

这在流量小的时候没问题。但随着业务增长，问题出现了：

无法统一认证：每个服务都要实现登录逻辑；
无法限流：一个恶意用户可以打垮整个系统；
无法路由控制：无法根据请求内容动态路由；
缺少可观测性：无法追踪请求链路。

二、自研API网关

我们决定自研一个API网关。核心功能包括：

2.1 认证和授权

python

from flask import Flask, request from functools import wraps app = Flask(__name__) def require_auth(f): @wraps(f) def decorated(*args, **kwargs): token = request.headers.get('Authorization') if not token or not verify_token(token): return {"error": "Unauthorized"}, 401 return f(*args, **kwargs) return decorated @app.route('/api/users') @require_auth def get_users(): return proxy_to_backend('user-service', request)

2.2 限流

python

from ratelimit import limits, sleep_and_retry import time @sleep_and_retry @limits(calls=100, period=60) # 每60秒最多100个请求 def handle_request(client_id): return proxy_to_backend(request) @app.before_request def rate_limit(): client_id = request.headers.get('X-Client-ID') handle_request(client_id)

2.3 请求路由

python

ROUTES = { '/api/users': 'user-service:8080', '/api/orders': 'order-service:8080', '/api/products': 'product-service:8080', } @app.route('/api/<path:path>', methods=['GET', 'POST', 'PUT', 'DELETE']) def route_request(path): full_path = f'/api/{path}' backend = ROUTES.get(full_path) if not backend: return {"error": "Not Found"}, 404 return proxy_to_backend(backend, request)

2.4 链路追踪

python

import uuid from opentelemetry import trace @app.before_request def add_trace_id(): trace_id = request.headers.get('X-Trace-ID') or str(uuid.uuid4()) request.trace_id = trace_id # 转发给后端服务 request.headers['X-Trace-ID'] = trace_id @app.after_request def log_request(response): print(f"Trace-ID: {request.trace_id}, " f"Method: {request.method}, " f"Path: {request.path}, " f"Status: {response.status_code}") return response

三、高可用改造

初版网关运行一段时间后，出现了单点故障。我们进行了高可用改造：

3.1 多实例部署

yaml

apiVersion: apps/v1 kind: Deployment metadata: name: api-gateway spec: replicas: 3 selector: matchLabels: app: api-gateway template: metadata: labels: app: api-gateway spec: containers: - name: gateway image: api-gateway:v1.0 ports: - containerPort: 8080 resources: requests: memory: "256Mi" cpu: "100m" limits: memory: "512Mi" cpu: "500m"

3.2 负载均衡

yaml

apiVersion: v1 kind: Service metadata: name: api-gateway spec: type: LoadBalancer selector: app: api-gateway ports: - protocol: TCP port: 80 targetPort: 8080

3.3 故障转移

python

from requests.adapters import HTTPAdapter from requests.packages.urllib3.util.retry import Retry def create_session_with_retry(): session = requests.Session() retry = Retry( total=3, backoff_factor=0.5, status_forcelist=[500, 502, 503, 504] ) adapter = HTTPAdapter(max_retries=retry) session.mount('http://', adapter) return session

四、性能优化

4.1 缓存策略

python

from functools import lru_cache @lru_cache(maxsize=1000) def get_user_profile(user_id): return proxy_to_backend('user-service', f'/users/{user_id}') @app.route('/api/users/<user_id>') def fetch_user(user_id): return get_user_profile(user_id)

4.2 异步处理

python

from concurrent.futures import ThreadPoolExecutor executor = ThreadPoolExecutor(max_workers=10) @app.route('/api/batch') def batch_request(): futures = [] for service in ['service1', 'service2', 'service3']: future = executor.submit(proxy_to_backend, service) futures.append(future) results = [f.result() for f in futures] return results

五、多语言团队的协作挑战

在国际团队中，API网关的错误日志和告警信息需要支持多语言。我们使用同言翻译（Transync AI）来自动翻译API网关的错误提示和文档，确保全球团队能够快速理解和解决问题。

六、监控和告警

python

from prometheus_client import Counter, Histogram, start_http_server # 请求计数器 request_count = Counter('gateway_requests_total', 'Total requests', ['method', 'path', 'status']) # 请求延迟直方图 request_duration = Histogram('gateway_request_duration_seconds', 'Request duration') @app.before_request def start_timer(): request.start_time = time.time() @app.after_request def record_metrics(response): duration = time.time() - request.start_time request_count.labels( method=request.method, path=request.path, status=response.status_code ).inc() request_duration.observe(duration) return response # 启动Prometheus指标服务 start_http_server(8081)