从Kratos设计看Go微服务工程实践

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"导读"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"github.com\/go-kratos\/kratos（以下简称Kratos）是一套轻量级 Go 微服务框架，致力于提供完整的微服务研发体验，整合相关框架及周边工具后，微服务治理相关部分可对整体业务开发周期无感，从而更加聚焦于业务交付。Kratos在设计之初就考虑到了高可扩展性，组件化，工程化，规范化等。对每位开发者而言，整套 Kratos 框架也是不错的学习仓库，可以了解和参考微服务的技术积累和经验。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"接下来我们从 "},{"type":"text","marks":[{"type":"strong"}],"text":"Protobuf"},{"type":"text","text":" 、 "},{"type":"text","marks":[{"type":"strong"}],"text":"开放性"},{"type":"text","text":" 、 "},{"type":"text","marks":[{"type":"strong"}],"text":"规范"},{"type":"text","text":" 、 "},{"type":"text","marks":[{"type":"strong"}],"text":"依赖注入"},{"type":"text","text":" 这4个点了解一下Kratos 在Go微服务工程领域的实践。"}]}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"基于Protocol Buffers(Protobuf)的生态"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在Kratos中，API定义、gRPC Service、HTTP Service、请求参数校验、错误定义、Swagger API json、应用服务模版等都是基于Protobuf IDL来构建的："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/9d\/86\/9d63d394e6d47f8f44322d1af42ce986.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"举一个简单的helloworld.proto例子："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"syntax = \"proto3\";\n\npackage helloworld;\n\nimport \"google\/api\/annotations.proto\";\nimport \"protoc-gen-openapiv2\/options\/annotations.proto\";\nimport \"validate\/validate.proto\";\nimport \"errors\/errors.proto\";\n\noption go_package = \"github.com\/go-kratos\/kratos\/examples\/helloworld\/helloworld\";\n\n\/\/ The greeting service definition.\nservice Greeter {\n\/\/ Sends a greeting\n rpc SayHello (HelloRequest) returns (HelloReply) {\n option (google.api.http) = {\n\/\/ 定义一个HTTP GET 接口，并且把 name 映射到 HelloRequest\nget: \"\/helloworld\/{name}\",\n };\n\/\/ 添加API接口描述(swagger api)\noption (grpc.gateway.protoc_gen_openapiv2.options.openapiv2_operation) = {\ndescription: \"这是SayHello接口\";\n };\n }\n}\n\n\/\/ The request message containing the user's name.\nmessage HelloRequest {\n\/\/ 增加name字段参数校验，字符数需在1到16之间\n string name = 1 [(validate.rules).string = {min_len: 1, max_len: 16}];\n}\n\n\/\/ The response message containing the greetings\nmessage HelloReply {\n string message = 1;\n}\n\nenum ErrorReason {\n\/\/ 设置缺省错误码\n option (errors.default_code) = 500;\n\/\/ 为某个错误枚举单独设置错误码\n USER_NOT_FOUND = 0 [(errors.code) = 404];\n CONTENT_MISSING = 1 [(errors.code) = 400];;\n}\n"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以上是一个简单的helloworld服务定义的例子，这里我们定义了一个Service叫Greeter，给Greeter添加了一个SayHello的接口，并根据googleapis规范给这个接口添加了Restful风格的HTTP接口定义，然后还利用openapiv2添加了接口的Swagger API描述，同时还给请求消息结构体HelloRequest中的name字段加上了参数校验，最后我们在文件的末尾还定义了这个服务可能返回的错误码。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这时我们在终端中执行:kratos proto client api\/helloworld\/ helloworld.proto 便可以生成以下文件："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/a3\/4d\/a3ayyfd9a54d5386d7691e3b068c3a4d.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"由上，我们看到Kraots脚手架工具帮我们一键生成了上面提到的能力。从这个例子中，我们可以直观感受到使用使用Protobuf带来的开发效率的提升，除此之外Kratos还有以下优点："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"清晰"},{"type":"text","text":"：做到了定义即文档，定义即代码"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"收敛，统一"},{"type":"text","text":"：将逻辑都收敛统一到一起，通过代码生成工具来保证HTTP Service、grpc Service等功能具有一致的行为"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"跨语言"},{"type":"text","text":"：众所周知Protobuf是跨语言的，java、go、python、php、js、c等等主流语言都支持"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"拥抱开源生态"},{"type":"text","text":"：比如Kratos复用了google.http.api、protoc-gen-openapiv2、protoc-gen-validate 等等一些犀利的Protobuf周边生态工具或规范，这比起自己造一个IDL的轮子要容易维护得多，同时老的使用这些轮子的gRPC项目迁移成本也更低"}]}]}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"开放性"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一个基础框架在设计的时候就要考虑未来的可扩展性，那Kratos是怎么做的呢？"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"1. Server Transport"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们先看下服务协议层的代码："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/9e\/87\/9e1db6d72e07a001f4bb5318d13f9887.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"上面是Kratos RPC服务协议层的接口定义，这里我们可以看到如果想要给Kratos新增一个新的服务协议，只要实现Start()、Stop()、Endpoint()这几个方法即可。这样的设计解耦了应用和服务协议层的实现，使得扩展服务协议更加方便。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/2e\/f4\/2ed441b60c8c33bc325f730986ed8ef4.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"从上图中我们可以看到App层无需关心底层服务协议的实现，只是一个容器管理好应用配置、服务生命周期、加载顺序即可。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"2. Log"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们再看一个Kratos日志模块的设计："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/5c\/e9\/5cb6a9a6aecb9f6cfbcb646d00e7f6e9.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这里Kratos定义了一个日志输出接口Logger，它的设计的非常简单 - 只用了一个方法、两个输入、一个输出。我们知道一个包暴露的接口越少，越容易维护，同时对使用和实现方的心智负担更小，扩展日志实现会变得更容易。但问题来了，这个接口从功能上来讲似乎只能输出日志level和固定的kv paris，如何能支持更高级的功能？比如输出 caller stack、实时timestamp、 context traceID ？这里我们定义了一个回调接口Valuer："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/f4\/7a\/f4f03588fb24c2c7a05c25e5f0fc2b7a.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这个Valuer可以被当作key\/value pairs中的value被Append到日志里，并被实时调用。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们看一下如何给日志加时间戳的Valuer实现："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/ca\/yy\/ca26644737503a508c30247d7769d9yy.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"使用时只要在原始的logger上再append一个固定的key和一个动态的valuer即可："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/74\/dd\/74bfe991b4033de582d056d29707fbdd.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这里的With是一个Helper function，里面new了一个新的logger（也实现了Logger接口），并将key\\value pairs暂存在新的logger里，等到Log方法被调用时再通过断言.(Valuer)的方式获取值并输出给底层原始的logger。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所以我们可以看到仅仅通过两个简单的接口+一个Helper function的组合我们就实现了日志的大多数功能，这样大大提高了可扩展性。实际上还有日志过滤、多日志源输出等功能也是通过组合使用这两接口来实现，这里待下次分享再展开细讲。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"3. Tracing"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"最后我们来看下Kratos的Tracing组件，这里Kratos采用的是CNCF项目OpenTelemetry。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"OpenTelemetry在设计之初就考虑到了组件化和高可扩展性，其实现了OpenTracing和W3C Trace Context的规范，可以无缝对接zipkin、jaeger等主流开源tracing系统，并且可以自定义Propagator 和 TraceProvider。通过otel.SetTracerProvider()我们可以轻易得替换Span的落地协议和格式，从而兼容老系统中的trace采集agent；通过otel.SetTextMapPropagtor()我们可以替换Span在RPC中的Encoding协议，从而可以和老系统中的服务互相调用时也能兼容。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"工程流程"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们知道在工程实践的时候，强规范和约束往往比自由和更多的选择更有优势，那么在Go工程规范这块我这里主要介绍三块："}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"1. 面向包的设计规范"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Go 是一个面向包名设计的语言，Package 在 Go 程序中主要起到功能隔离的作用，标准库就是很好的设计范例。Kratos也是可以按包进行组织代码结构，这里我们抽取Kratos根目录下主要几个Package包来看下："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"\/cmd"},{"type":"text","text":"：可以通过 go install 一键安装生成工具，使用户更加方便地使用框架。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"\/api"},{"type":"text","text":"：Kratos框架本身的暴露的接口定义"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"\/errors"},{"type":"text","text":"：统一的业务错误封装，方便返回错误码和业务原因。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"\/config"},{"type":"text","text":"：支持多数据源方式，进行配置合并铺平，通过 Atomic 方式支热更配置。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"\/internal"},{"type":"text","text":" ：存放对外不可见或者不稳定的接口。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"\/transport"},{"type":"text","text":"：服务协议层（HTTP\/gRPC）的抽象封装，可以方便获取对应的接口信息。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"\/middleware"},{"type":"text","text":"：中间件抽象接口，主要跟transport 和 service 之间的桥梁适配器。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"\/third_party"},{"type":"text","text":"：第三方外部的依赖"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"可以看到Kratos的包命名清晰简短，按功能进行划分，每个包具有唯一的职责。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在设计包时我们还需要考虑到以下几点："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"包的设计必须以使用者为中心，直观且易于使用，包的命名必须旨在描述它提供的内容，如果包的名称不能立即暗示这一点，则它可能包含一组零散的功能。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"包的目的是为特定问题域而提供的，为了有目的，包必须提供，而不是包含。包不能成为不同问题域的聚合地，随着时间的推移，它将影响项目的简洁和重构、适应、扩展和分离的能力。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"高便携性，尽量减少依赖其他代码库，一个包与其它包依赖越少，一个包的可重用性就越高。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"不能成为单点依赖，当包被单一的依赖点时，就像一个公共包（common），会给项目带来很高的耦合性。"}]}]}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"2. 配置"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"首先，我们来看下常见的基础框架是怎么初始化配置的："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/59\/1c\/598da73d16860f4efa621dfb329e291c.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这是Go标准库HTTP Server配置初始化的例子，但是这样做会有如下几个问题："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"&http.Server{}由于是一个取址引用，里面的参数可能会被外部运行时修改，这种运行时修改带来的危害是不可把控的。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"无法区分nil和0值，当里面的参数值为0的时候，不知道是用户未设置还是就是被设置成了0。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"难以分辨必传和选传参数，只能通过文档说明来隐式约定，没有强约束力。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"那么Kraots是怎么解决这些问题的呢？答案就是Functional Options 。我们看下transport\/http\/client.go的代码："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/9d\/cf\/9d5c4bed31d89e1e211f9734b74947cf.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Client.go中定义了一个回调函数ClientOption，该函数接受一个定义了一个存放实际配置的未导出结构体clientOptions的指针，然后我们在NewClient的时候，使用可变参数进行传递，然后再初始化函数内部通过 for 循环调用修改相关的配置。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这么做有这么几个好处："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"由于clientOptions结构体是未导出的，那么就不存在被外部修改的可能。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"可以区分0值和未设置，首先我们在new clientOptions时会设置默认参数，那么如果外部没有传递相应的Option就不会修改这个默认参数。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"必选参数显示定义，可选值则通过Go可变参数进行传递，很好的区分必传和选传。"}]}]}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"3. Error规范"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Kratos为微服务提供了统一的Error模型："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/5c\/4f\/5cae8a9144188ccb20edf32544256f4f.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Code用作外部展示和初步判断，服务端无需定义大量全局唯一的XXX_NOT_FOUND，而是使用一个标准Code.NOT_FOUND错误代码并告诉客户端找不到某个资源。错误空间变小降低了文档的复杂性，在客户端库中提供了更好的惯用映射，并降低了客户端的逻辑复杂性。同时这种标准的大类Code的存在也对外部的观测系统更友好，比如可以通过分析Nginx Access Log中的HTTP StatusCode来做服务端监控和告警。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Reason是具体的错误原因，可以用来更详细的错误判定。每个微服务都会定义自己Reason，那么要保持全局唯一就需要加上领域前缀，比如User_XXX。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Message错误信息可以帮助用户轻松快捷地理解和解决API 错误"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Metadata中则可以存放一些标准的错误详情，比如retryInfo、error stack等"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这种强制规范，避免了开发人员直接透传Go的error 从而导致一些敏感信息泄露。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"接下来我们看下Error结构体还实现了哪些接口："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/36\/ff\/36909c8d300b8c0901fdbdd6ffd8b5ff.jpg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"实现了GRPCStatus () *status.Status 接口，这样就实现了从http status code到grpc status code的转换，这样Kratos Error可以被gRPC直接转成google.rpc.Status传递出去。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"实现了标准库errors包的Is (error) bool接口，这样使用者可以直接调用errors.Is()来比较两个erorr中的reason是否相等，避免了使用==来直接判断error是否相等这种错误姿势。"}]}]}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"依赖注入"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"依赖注入 (Dependency Injection)可以理解为一种代码的构造模式，按照这样的方式来写，能够让你的代码更加容易维护，一般在Java的项目中见到的比较多。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"依赖注入初看起来比较违反直觉，那么为什么Go也需要依赖注入？假设我们要实现一个用户访问计数的功能。我们先看看不使用依赖注入的项目代码："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"type Service struct {\n redisCli *redis.Client\n}\n\nfunc (s *Service) AddUserCount(ctx context.Context) {\n \/\/do some business logic\n s.redisCli.Incr(ctx, \"user_count\")\n}\n\nfunc NewService(cfg *redis.Options) *Service {\n return &Service{redisCli: redis.NewClient(cfg)}\n}\n"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这种方式比较常见，在项目刚开始或者规模小的时候没什么问题，但我们如果考虑下面这些因素："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Redis是基础组件，往往会在项目的很多地方被依赖，那么如果哪天我们想整体修改redis sdk的甚至想把redis 整体替换成mysql时，需要在每个被用到的地方都进行修改，耗时耗力还容易出错。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"很难对App这个类写单元测试，因为我们需要创建一个真实的redis.Client。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"使用依赖注入改造后的Service："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"type DataSource interface{\n Incr(context.Context, string)\n}\n\ntype Service struct {\n dataSource DataSource\n}\n\nfunc (s *Service) AddUserCount(ctx context.Context) {\n \/\/do some business logic\n s.dataSource.Incr(ctx, \"user_count\")\n}\n\nfunc NewService(ds DataSource) *Service {\n return &Service{dataSource: ds}\n}\n"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"上面代码中我们把*redis.Client实体替换成了一个DataSource接口，同时不控制dataSource的创建和销毁，把dataSource生命周期控制权交给了上层来处理，以上操作有三个主要原因："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"因为Service层已不再关心dataSource的创建和销毁，这样当我们需要修改dataSource实现的时候，只要在上层统一修改即可，无需在各个被依赖的地方一一修改。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"因为依赖的是一个接口，我们写单元测试的时候只要传递一个mock后的Datasource实现即可 。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这里dataSource这个基础组件不再被会到处创建，可以做到复用一个单例节省资源开销。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Go 的依赖注入框架有两类，一类是通过反射在运行时进行依赖注入，典型代表是 uber 开源的 dig，另外一类是通过 generate 进行代码生成，典型代表是 Google 开源的 wire。使用 dig 功能会强大一些，但是缺点就是错误只能在运行时才能发现，这样如果不小心的话可能会导致一些隐藏的 bug 出现。使用 wire 的缺点就是功能限制多一些，但是好处就是编译的时候就可以发现问题，并且生成的代码其实和我们自己手写相关代码差不太多，更符合直觉，心智负担更小。所以Kratos更加推荐 wire，Kratos的默认项目模板中 kratos-layout 也正是使用了 google\/wire 进行依赖注入。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们来看下wire使用方式："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们首先要定义一个ProviderSet，这个Set会返回构建依赖关系所需的组件Provider。如下所示，Provider往往是一些简单的工厂函数，这些函数不会太复杂："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"type RedisSource struct {\n redisCli *redis.Client\n}\n\n\/\/ RedisSource实现了Datasource的Incr接口\nfunc (ds *RedisSource) Incr(ctx context.Context, key string) {\n ds.redisCli.Incr(ctx, key)\n}\n\n\/\/ 构建实现了DataSource接口的Provider\nfunc NewRedisSource(db *redis.Client) *RedisSource {\n return &RedisSource{redisCli: db}\n}\n\n\/\/ 构建*redis.Client的Provider\nfunc NewRedis(cfg *redis.Options) *redis.Client {\n return redis.NewClient(cfg)\n}\n\/\/ 这是一个Provider的集合，告诉wire这个包提供了哪些Provider\nvar ProviderSet = wire.NewSet(NewRedis, NewRedisSource)\n"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"接着我们要在应用启动处新建一个wire.go文件并定义Injector，Injctor会分析依赖关系并将Provider串联起来构建出最终的Service："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"\/\/ +build wireinject\n\nfunc initService(cfg *redis.Options) *service.Service {\n panic(wire.Build(\n redisSource.ProviderSet,\n\/\/使用 wire.Bind 将 Struct 和接口进行绑定了，表示这个结构体实现了这个接口，\nwire.Bind(new(data.DataSource), new(*redisSource.RedisSource)),\n service.NewService),\n )\n}\n"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"最后执行wire .后自动生成的代码如下："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"\/\/go:generate go run github.com\/google\/wire\/cmd\/wire\n\/\/+build !wireinject\n\nfunc initService(cfg *redis.Options) *service.Service {\n client := redis2.NewRedis(cfg)\n redisSource := redis2.NewRedisSource(client)\n serviceService := service.NewService(redisSource)\n return serviceService\n}\n"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"由此我们可以看到只要定义好组件初始化的Provider函数，还有把这些Provider组装在一起的Injector就可以直接生成初始化链路代码了，上手还是相对简单的，生成的代码所见即所得，容易Debug。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"综上可见，Kratos是一款凝结了开源社区力量以及Go同学们大量微服务工程实践后诞生的一款微服务框架。现在腾讯云微服务治理治理平台(微服务平台TSF)也已支持Kratos框架，给Kratos赋予了更多企业级服务治理能力、提供多维度服务，如：应用生命周期托管、一键上云、私有化部署、多语言发布。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"作者介绍"},{"type":"text","text":"："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"曹国梁"},{"type":"text","text":"：6年Go微服务研发经历，腾讯云高级研发工程师，Kratos Maintainer，gRPC-go contributor"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"本文转载自公众号腾讯云中间件（ID：gh_6ea1bc2dd5fd）。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"原文链接"},{"type":"text","text":"："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/mp.weixin.qq.com\/s\/ZCoc3qA_r1fKzPOURA2V2g","title":"","type":null},"content":[{"type":"text","text":"从Kratos设计看Go微服务工程实践"}]}]}]}