[learn microservices from 0] [03] explore the microservice architecture

---

Continuous learning & Ongoing update …

Keep breaking away

---

A normal service invocation process

• First, the service provider （ It's the service provider ） In a certain format The service description , towards Registry Center Registration service , State what services you can provide and what the address of the service is , Complete service release .

• Next, serve consumers （ It's the party that invokes the service ） Ask the registry , Query the address of the service to be called , Then, a request is made to the service provider with the agreed communication protocol , After getting the request result, parse the result according to the agreed protocol .

• And during the invocation of the service , Service requests take time 、 The call volume, success rate and other indicators will be recorded as monitor , The link information of the call will be recorded , be used for Fault location and problem tracking . in the meantime , If the call fails , You can try again Service governance Means to ensure the success rate .

• To sum up , Microservices architecture , Service invocation relies on the following basic components ： The service description 、 Registry Center 、 Service Framework 、 Service monitoring 、 Service tracking 、 Service governance

The service description

• The first problem to be solved by service invocation is how to describe the service externally . such as , You have provided a service , What is the service name of this service ？ What information needs to be provided to call this service ？ What format is the result returned by calling this service ？ How to analyze ？ These are the problems to be solved by the service description .

• Common service descriptions include RESTful API、XML The configuration and IDL There are three kinds of documents .

• among ,RESTful API The method is usually used for HTTP Service description of the agreement , And often Wiki perhaps Swagger To manage .

• XML Configuration mode is mostly used as RPC Service description of the agreement , adopt *.xml Configuration file to define the interface name 、 Parameters and return value types .

• IDL File mode is usually used for Thrift and gRPC In this kind of cross language service invocation framework , such as gRPC It is through Protobuf File to define the interface name of the service 、 Parameters and the data structure of the return value .

Registry Center

• With the interface description of the service , The next step is to solve the problem of service publishing and subscription , That means you provide a service , How to let external people who want to call your service know . At this time, you need a role similar to the registry , Service providers register their services and addresses with the registry , The service consumer queries the registry for the address of the service to be called , Then make a request .

• In general , The workflow of the registry is ：

  • The service provider at startup , Register your own service with the registry according to the publishing information configured in the service publishing file .

  • Service consumers at startup , According to the service information configured in the consumer profile, subscribe to the registry for the services you need .

  • The registry returns the service provider address list to the service consumer .

  • When the service provider changes , For example, nodes are added or destroyed , The registry notifies the service consumer of the change .

Service Framework

• Through the registry , The service consumer can get the address of the service provider , When you have an address, you can initiate a call . But before you start the call, you need to solve the following problems .

• What protocol is used for service communication ？ That is to say, what kind of protocol is used for network communication between service provider and service consumer , It's four layers TCP、UDP agreement , Or seven layers HTTP agreement , Or other agreements ？

• What is the method of data transmission ？ That is to say, which method is adopted for data transmission between service provider and service consumer , Synchronous or asynchronous , It's on a single connection , Or multiplexing .

• What format does data compression use ？ Data transmission usually compresses data , To reduce the amount of data transmitted over the network , Thus reducing bandwidth consumption and network transmission time , For example, the common JSON serialize 、Java Object serialization and Protobuf Serialization, etc .

• …

• This is the work of the service framework

Service monitoring

• Once the service consumer and service provider can normally initiate service invocation , You need to monitor the call , To find out if the service is normal . Generally speaking , Service monitoring mainly includes three processes .

• Index collection . It is to collect the time-consuming and successful requests of each service call , And upload it to the centralized data processing center .

• Data processing . With the time-consuming and successful information of each call , You can calculate the number of service requests per second 、 Average time and success rate .

• Data presentation . Data collection , After processing , It also needs to be displayed in a friendly way , In order to play a value . The data is usually displayed in Dashboard Panel , And every 10s Auto refresh at equal intervals , Used for business monitoring and alarm etc .

Service tracking

• In addition to the need to monitor service invocation , You also need to record every link that the service calls through , For problem tracking and fault location .

• The working principle of service tracking is as follows ：

• Before the service consumer initiates the call , Will generate a... Locally according to certain rules requestid, When the call is initiated , take requestid As part of the request parameters , Pass it on to the service provider .

• After the service provider receives the request , Record the requested requestid, Then process the request . If the service provider continues to request other services , It will generate its own requestid, Then put these two requestid All continue to pass down as request parameters .

• And so on , Through this layer by layer transmission , One request , No matter how many service calls it depends on finally 、 How many service nodes have passed , Can be generated through the first requestid Concatenate all nodes , So as to achieve the purpose of service tracking .

Service governance

• Service monitoring can find problems , Service tracking can locate the problem , And solving the problem depends on Service Management . Service governance is to ensure that in all kinds of unexpected situations , Service calls still work .

• In the production environment , You should often encounter the following situations .

• Single machine failure . Usually encounter single machine failure , All rely on operation and maintenance to find and restart services or remove faulty nodes from the line . However, the larger the cluster , The more likely it is to encounter single machine failure , When the scale of the machine is more than 100 , It is obviously difficult to cope with the traditional human flesh operation and maintenance . Service governance can be achieved through certain strategies , Automatic removal of faulty nodes , There is no need for human intervention , It can ensure that single machine failure will not affect the business .

• single IDC fault . You should often hear so and so App, Because of the serious failure that a large number of users cannot use due to the excavation of optical cables during construction . Service governance can be achieved through automatic failover IDC The flow to other normal IDC, Can avoid because of single IDC A large number of businesses affected by the failure .

• Dependent services are not available . For example, your service depends on another service , When another service has a problem , It will slow down or even bring down your service . And service governance can be achieved through fusing , In the case of a dependency service exception , Stop calling for a period of time and return to . On the one hand, this ensures that service consumers will not be dragged down , On the other hand, it also reduces the pressure on service providers , To enable it to recover as soon as possible .

• The above are the three most common scenarios that require the introduction of service governance , Of course, there are other means of service governance, such as automatic capacity expansion and contraction , It can be used to solve the problem of service capacity .

summary

• These basic components together form a micro service architecture , In the production environment, it is indispensable
• So before introducing microservice architecture , Your team must master the principles of these basic components and have the corresponding development ability .
• In the way of realization , Open source solutions can be introduced ; If there are enough senior technicians , You can also choose to develop each component of the microservice architecture by yourself . But for most small and medium-sized teams , I think using open source implementation is a wiser choice , On the one hand, you can save the investment of relevant technicians and focus more on the business , On the other hand, you can also take fewer detours and step on fewer pits .
• Whether you use open source solutions or develop your own , Must understand the working principle of each component and be able to carry out secondary development on this basis .

Reference resources

Geek time : from 0 Start learning about microservices -- First line practical experience of microblog service experts

---

The end of this paper , Thank you for your attention and support ！

---