自研企业级API网关 在现代的微服务架构中，网关是至关重要的组件，它不仅起到路由的作用，还需要处理各种复杂的网络问题如用户的权限认证、限流、降级熔断、灰度发布负载均衡等。Spring Cloud Gateway 作为市场上的佼佼者，已经为众多企业提供了稳定和可靠的服务。但是，每个企业和场景都有其独特性，并且Spring Cloud Gateway对于中小型企业来说用起来 “太重” 了。 在这样的环境下，诞生了自研的企业级网关。与 Spring Cloud Gateway 相比，我们的网关更加灵活和轻量，特别优化了在特定场景和负载下的性能。同时，我们也引入了一些创新的功能和策略，以满足企业级应用中的特定需求。
Netty + Nacos + Disruptor + grafana + prometheus。考虑到轻量和功能，目前这几个技术栈是足够的，后续如果要继续完善功能点，可以适当加入其他技术栈。这也体现了自研网关的优点——可定制化。
- 使用了 Disruptor 框架，其是一个高性能的、无锁的环形数组结构，主要用于实现日志框架、消息中间件等场景的高性能异步处理。它被设计出来是为了替代某些使用队列的场景，并提高吞吐量。在该自研网关中是用来接受客户端请求，然后经过消费者消费转发到具体后端服务。经过jmeter测压，其与传统的阻塞队列（如Java的 ArrayBlockingQueue）相比，吞吐量可以提高10倍甚至更多。
- 引入了Caffeine缓存库，具有灵活的缓存策略、独特的刷新机制，同时也是无锁设计的。
- 通过 grafana + prometheus 展示流量（请求）的相关属性，便于对流量进行分析，从而调整相关限流参数等等。
- 路由
- 负载均衡
- 限流
- 灰调发布
- 请求重试
- 降级熔断
- 请求的流量监控
- 用户的权限认证
前提说明：该自研网关中，自定义了 @ApiInvoker 和 @ApiService这两个注解。具体如何使用看演示模块代码（backend-user-service或者backend-http-server）。
- 路由
将打到网关的客户端请求转发到后端的具体服务。
{ "rules": [ { "id": "1", "name": "test-1", "protocol": "http", "serviceId": "backend-http-server", "prefix": "/user", "paths": [ "/http-server/ping", "/user/update" ], "filterConfigs": [ { "id": "router_filter" } ] } ] } - 负载均衡
目前实现了基于轮询和随机的负载均衡。
{ "rules": [ { "id": "1", "name": "test-1", "protocol": "http", "serviceId": "backend-http-server", "prefix": "/user", "paths": [ "/http-server/ping", "/user/update" ], "filterConfigs": [ { "id": "load_balance_filter", "config": { "load_balance": "RoundRobin" } }, { "id": "router_filter" } ] } ] } - 限流
目前具有基于redis实现的分布式限流器和基于Guava RateLimiter实现的单机限流器。
{ "rules": [ { "id": "1", "name": "test-1", "protocol": "http", "serviceId": "backend-http-server", "prefix": "/user", "paths": [ "/http-server/ping", "/user/update" ], "filterConfigs": [ { "id": "load_balance_filter", "config": { "load_balance": "RoundRobin" } }, { "id": "router_filter" }, { "id": "flow_ctl_filter" } ], "flowCtlConfigs": [ { "type": "path", "model": "distributed", "value": "/http-server/ping", "config": { "duration": 10, "permits": 20, "maxPermits": 14 } } ] } ] } - 灰度发布
哪个模块被定义为灰度发布模块就在该模块的resource的配置文件里写入 api.gray=true。以此对应的，哪个请求要被转发到灰度发布的模块就必须添加额外请求头——gray:true。
{ "rules": [ { "id": "1", "name": "test-1", "protocol": "http", "serviceId": "backend-http-server", "prefix": "/user", "paths": [ "/http-server/ping", "/user/update" ], "filterConfigs": [ { "id": "load_balance_filter", "config": { "load_balance": "RoundRobin" } }, { "id": "router_filter" }, { "id": "flow_ctl_filter" }, { "id":"gray_filter" } ], "flowCtlConfigs": [ { "type": "path", "model": "distributed", "value": "/http-server/ping", "config": { "duration": 10, "permits": 20, "maxPermits": 14 } } ] } ] } - 请求重试
目前只实现了简单的请求重试机制，后续可以引入一些请求重试框架如Failsafe、Resilience4j等对该功能进一步完善。
{ "rules": [ { "id": "1", "name": "test-1", "protocol": "http", "serviceId": "backend-http-server", "prefix": "/user", "paths": [ "/http-server/ping", "/user/update" ], "filterConfigs": [ { "id": "load_balance_filter", "config": { "load_balance": "RoundRobin" } }, { "id": "router_filter" }, { "id": "flow_ctl_filter" }, { "id":"gray_filter" } ], "flowCtlConfigs": [ { "type": "path", "model": "distributed", "value": "/http-server/ping", "config": { "duration": 10, "permits": 20, "maxPermits": 14 } } ], "retryConfig": { "times": 5 } } ] } - 降级熔断
这个降级熔断是通过hystrix实现的，这部分功能很强大，感兴趣的可以好好研究并且进一步扩展。
{ "rules": [ { "id": "1", "name": "test-1", "protocol": "http", "serviceId": "backend-http-server", "prefix": "/user", "paths": [ "/http-server/ping", "/user/update" ], "filterConfigs": [ { "id": "load_balance_filter", "config": { "load_balance": "RoundRobin" } }, { "id": "flow_ctl_filter" }, { "id": "router_filter" }, { "id": "gray_filter" } ], "flowCtlConfigs": [ { "type": "path", "model": "distributed", "value": "/http-server/ping", "config": { "duration": 10, "permits": 20, "maxPermits": 14 } } ], "retryConfig": { "times": 5 }, "hystrixConfigs": [ { "path": "/http-server/ping", "threadCoreSize": 2, "timeoutInMilliseconds": 1000, "timeoutEnabled": true, "slidingWindowDuration": 30000, "numberOfWindowSegments": 10, "requestThreshold": 4, "failureRateThreshold": 50, "circuitBreakerResetTime": 30000, "circuitBreakerEnabled": true, "fallbackResponse": "熔断时间", "globalEnable": false } ] } ] } - 请求的流量监控
这里通过添加两个过滤器（利用prometheus收集数据）实现对请求流量的监控，并且最后通过定制化面板在grafana展示出来。
{ "rules": [ { "id": "1", "name": "test-1", "protocol": "http", "serviceId": "backend-http-server", "prefix": "/user", "paths": [ "/http-server/ping", "/user/update" ], "filterConfigs": [ { "id": "load_balance_filter", "config": { "load_balance": "RoundRobin" } }, { "id": "flow_ctl_filter" }, { "id": "router_filter" }, { "id": "gray_filter" }, { "id": "monitor_filter" }, { "id": "monitor_end_filter" } ], "flowCtlConfigs": [ { "type": "path", "model": "distributed", "value": "/http-server/ping", "config": { "duration": 10, "permits": 20, "maxPermits": 14 } } ], "retryConfig": { "times": 5 }, "hystrixConfigs": [ { "path": "/http-server/ping", "threadCoreSize": 2, "timeoutInMilliseconds": 1000, "timeoutEnabled": true, "slidingWindowDuration": 30000, "numberOfWindowSegments": 10, "requestThreshold": 4, "failureRateThreshold": 50, "circuitBreakerResetTime": 30000, "circuitBreakerEnabled": true, "fallbackResponse": "熔断时间", "globalEnable": false } ] } ] } - 用户的权限认证
{ "rules": [ { "id": "2", "name": "user-private", "protocol": "http", "serviceId": "backend-user-server", "prefix": "/user/private", "paths": [ "/user/private/user-info" ], "filterConfigs": [ { "id": "load_balance_filter", "config": { "load_balance": "RoundRobin" } }, { "id": "user_auth_filter" }, { "id": "router_filter" } ] } ] }