GitHub - zzhujing/spring-cloud-learning: SpringCloud+SpringCloud Alibaba 脚手架工程

SpringCloud

在复习SpringCloud+SpringCloud Alibaba时进行微服务各种组件的集成

分布式理论
服务注册发现
远程调用
服务容错、熔断、限流
路由网关
分布式配置中心
分布式消息服务
分布式链路追踪
分布式事务

CAP

Consistency ：一致性
Availabiity : 可用性
Parition Tolerance ：分区容错性

一般是CP或者AP，在分布式环境下必须实现P，如下注册中心，Eureka就是AP，ZK和Consul是CP。

一般AP最后会通过BASE理论来完成数据最终一致性。比如底层NOSQL的数据同步等。

1、服务注册与发现

Eureka

server

环境搭建

父pom

            <!--spring boot 2.2.2-->
            <dependency>
                <groupId>org.springframework.boot</groupId>
                <artifactId>spring-boot-dependencies</artifactId>
                <version>2.2.2.RELEASE</version>
                <type>pom</type>
                <scope>import</scope>
            </dependency>
            <!--spring cloud Hoxton.SR1-->
            <dependency>
                <groupId>org.springframework.cloud</groupId>
                <artifactId>spring-cloud-dependencies</artifactId>
                <version>Hoxton.SR1</version>
                <type>pom</type>
                <scope>import</scope>
            </dependency>

添加maven依赖

        
       <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-netflix-eureka-server</artifactId>
        </dependency>

添加配置

server:
  port: 7001
eureka:
  instance:
    hostname: eureka7001.com #服务端实例名
  client:
    register-with-eureka: false #不注册自己
    fetch-registry: false #不需要拉取服务
    service-url:
      defaultZone: http://eureka7002.com:7002/eureka/ #集群配置，配置另外一server节点的地址
  server:
    enable-self-preservation: false #关闭自我保护机制，开发过程请关闭！生产打开保证高可用
    eviction-interval-timer-in-ms: 2000 #清除非正常服务周期

添加注解@EnableEurekaServer，启动即可

client

maven依赖

       <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-netflix-eureka-client</artifactId>
        </dependency>

添加注解@EnableEurekaClient/@EnableDiscoveryClient
添加配置

      
eureka:
  client:
    service-url:
      defaultZone: http://eureka7001.com:7001/eureka/,http://eureka7002.com:7002/eureka/
    register-with-eureka: true #注册自己
    fetch-registry: true #拉取服务信息，集群必须配置才能配置ribbon实现负载均衡
  instance:
    prefer-ip-address: true #使用ip注册
    lease-expiration-duration-in-seconds: 2 #最短超时周期，超过2s则超时
    lease-renewal-interval-in-seconds: 1 #最短心跳发送周期

Zookeeper

server

使用docker启动zk

d run --name zk01 -d  -p2181:2181 -p2888:28888 -p 3888:3888 zookeeper

client

添加maven依赖

        <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-zookeeper-discovery</artifactId>
        </dependency>

配置yaml

spring:
  application:
    name: cloud-provider-payment
  cloud:
    zookeeper:
      connect-string: localhost:2181 #zk server地址

添加注解@EnableDiscoveryClient

Consul

Server

使用docker启动

d run --name consul -d -p8500:8500 -p8600:8600/udp consul

Client

添加maven依赖

        <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-consul-discovery</artifactId>
        </dependency>

添加配置

spring:
  application:
    name: cloud-consumer-order
  cloud:
    consul:
      host: localhost
      port: 8500
      discovery:
        prefer-ip-address: true

添加注解@EnableDiscoveryClient

Nacos

Server

使用官方的nacos包进行集群搭建

在对应的conf/application.properties下面添加自定义的持久化MySQL数据库的配置，并使用官方的SQL初始化对应的数据库和表

spring.datasource.platform=mysql
db.num=1 (MySQL配置的个数)
db.url.0=jdbc:mysql://127.0.0.1:13306/nacos_config?characterEncoding=utf8&connectTimeout=1000&socketTimeout=3000&autoReconnect=true
db.user=root
db.password=hujing19960825

修改conf/cluster.conf的IP配置

192.168.88.15:3333
192.168.88.15:4444
192.168.88.15:5555  #集群ip

到bin目录下修改startup.sh用来支持自定义端口

export FUNCTION_MODE="all"
while getopts ":m:f:s:p:" opt #这里添加了:p代表端口
do
    case $opt in
        m)
            MODE=$OPTARG;;
        f)
            FUNCTION_MODE=$OPTARG;;
        s)
            SERVER=$OPTARG;;
#######  添加内容     ############
        p)
            PORT=$OPTARG;;
#######  添加内容     ############
        ?)
        echo "Unknown parameter"
        exit 1;;
    esac
done
#最后执行的时候添加--server.port参数
nohup $JAVA ${JAVA_OPT} --server.port=${PORT} nacos.nacos >> ${BASE_DIR}/logs/start.out 2>&1 &

使用sh startup.sh -p 启动即可
配置nginx进行反向代理

    upstream cluster {
        server 192.168.1.20:8848;
        server 192.168.1.20:8849;
        server 192.168.1.20:8850;
    }
    server {
        listen       80;
        server_name  localhost;
        location / {
           proxy_pass http://cluster;
        }
   }

基于docker搭建Nacos-Server

修改docker-compose.yml

有下面几个注意点
- 官方默认MySQL是支持mysql5，这里使用数据卷将MySQL8.x的jar包设置到nacos容器中
- nacos版本使用的是1.2.0

version: "2"
services:
  nacos1:
    image: nacos/nacos-server:latest
    container_name: nacos1
    networks:
      nacos_net:
        ipv4_address: 172.16.238.10
    volumes:
      - ./plugins/mysql/:/home/nacos/plugins/mysql/
      - ./cluster-logs/nacos1:/home/nacos/logs
    ports:
      - "8848:8848"
      - "9555:9555"
    env_file:
      - ../env/nacos-ip.env
    restart: on-failure
  nacos2:
    image: nacos/nacos-server:latest
    container_name: nacos2
    networks:
      nacos_net:
        ipv4_address: 172.16.238.11
    volumes:
      - ./plugins/mysql/:/home/nacos/plugins/mysql/
      - ./cluster-logs/nacos2:/home/nacos/logs
    ports:
      - "8849:8848"
    env_file:
      - ../env/nacos-ip.env
    restart: always
  nacos3:
    image: nacos/nacos-server:latest
    container_name: nacos3
    networks:
      nacos_net:
        ipv4_address: 172.16.238.12
    volumes:
      - ./plugins/mysql/:/home/nacos/plugins/mysql/
      - ./cluster-logs/nacos2:/home/nacos/logs
    ports:
      - "8850:8848"
    env_file:
      - ../env/nacos-ip.env
    restart: always
networks:
  nacos_net:
    driver: bridge
    ipam:
      driver: default
      config:
        - subnet: 172.16.238.0/24

配置env/nacos-ip.yml

NACOS_SERVERS=172.16.238.10:8848 172.16.238.11:8848 172.16.238.12:8848 #设置docker集群地址
MYSQL_SERVICE_HOST=192.168.1.20 #数据库ip
MYSQL_SERVICE_DB_NAME=nacos_config #数据库名称
MYSQL_SERVICE_PORT=3307
MYSQL_SERVICE_USER=hujing
MYSQL_SERVICE_PASSWORD=hujing19960825

Client

maven配置

        <dependency>
            <groupId>com.alibaba.cloud</groupId>
            <artifactId>spring-cloud-starter-alibaba-nacos-discovery</artifactId>
        </dependency>

yml

spring:
  application:
    name: nacos-provider-payment
  cloud:
    nacos:
      server-addr: localhost:8848 # nacos服务地址

2、远程调用

Ribbon+RestTemplate

@Configuration
class Config{
  
  @LoadBalanced //开启ribbon负载均衡
  @Bean
  public RestTemplate restTemplate(){
    return new RestTemplate();
  }
}

可以通过@RibbonClient来扩展ribbon配置，但是不能被SpringBoot扫描到，否则会被Spring上下文扫描到变成全局配置

//该扩展配置文件不能被SpringBoot 扫描到，不然会被Spring上下文扫描到，导致所有的实例名(当前配置对应的实例和其他所有实例都包括)
// 都会被该默认配置覆盖。
@Configuration
public class CustomRibbonConfig {
    @Bean
    public IRule rule(){ //配置负载均衡策略
        return new RandomRule();
    }
}

在yaml中配置ribbon,详细配置项可以看CommonClientConfigKey

<service name>:
	ribbon:
		xxxx: yyyy

---

ribbon:
	ReadTimeout: 5000 //配置超时时间
	ConnectTimeout: 5000

Feign+Ribbon

@FeignClient(value = "cloud-provider-payment",configuartion="xxxFeign扩展配置")
public interface PaymentClient {

    @GetMapping("echo/{id}")
    String echo(@PathVariable("id") Integer id);

yaml配置全局feign配置

feign:
  client:
    config:
      default:
        logger-level: full #日志级别，可查看FeignClientConfiguration
        connect-timeout: 5000 #feignClient粒度的超时时间配置
        read-timeout: 5000
  hystrix:
    enabled: true #开启feign的hystrix支持
logging:
  level:
    cloud.study: debug

3、服务容错、熔断、限流

Hystrix

集成依赖

       <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-netflix-hystrix</artifactId>
        </dependency>

开启注解@EnableCircuitBreker

容错支持`@HystrixCommand`

    @GetMapping("/echoWithHystrix")
    @HystrixCommand(fallbackMethod = "echoFallback", commandProperties = {
            @HystrixProperty(name = "execution.isolation.thread.timeoutInMilliseconds", value = "3000") //设置熔断超时时间
    })
    public String echoWithHystrix() {
        try {
            TimeUnit.SECONDS.sleep(5);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        return "执行成功O(∩_∩)O哈哈~";
    }

    /**
     * Hystrix fallBack method
     * 1.超时
     * 2.异常报错
     */
    public String echoFallback() {
        return "服务器繁忙o(╥﹏╥)o";
    }

更多HystrixCommand属性可以参考HystrixCommandProperties

熔断，断路器模式

hystrix工作流程图

是否缓存，缓存直接返回
是否在断路开启状态，是直接返回fallback
不是则查看线程池/信号量是否满了
执行方法查看是否执行失败或者超时

![image-20200328175345561](/Users/hujing/Library/Application Support/typora-user-images/image-20200328175345561.png)

    @GetMapping("/echo1/{id}")
    @HystrixCommand(defaultFallback = "defaultFallback", commandProperties = {
        //开启断路器模式，默认true
        @HystrixProperty(name = "circuitBreaker.enabled", value = "true"),
        //请求到达多少次才会触发熔断，默认20次
        @HystrixProperty(name = "circuitBreaker.requestVolumeThreshold", value = "10"),
        //熔断触发时间窗口
        @HystrixProperty(name = "circuitBreaker.sleepWindowInMilliseconds", value = "30000"),
        //百分之10就进行熔断
        @HystrixProperty(name = "circuitBreaker.errorThresholdPercentage", value = "10")
    })
    public String echo1(@PathVariable Integer id) {
        return paymentClient.echo(id);
    }

在一段时间内达到请求次数，并且达到错误率则进入熔断状态
在一个窗口时间结束后，切换为半开状态，若下次请求还是失败，继续进入熔断
若下次请求成功则关闭熔断

解决太多熔断方法膨胀代码？

@DefaultProperties(defaultFallback = "defaultFallback") //公用fallback

解决fallback方法和业务耦合在一起

@FeignClient(value = "cloud-provider-payment" ,fallback = PaymentClient.PaymentClientFallback.class) //使用fallback,fallbackFactory来处理耦合
public interface PaymentClient {

    @GetMapping("echo/{id}")
    String echo(@PathVariable("id") Integer id);

		//可以防止到单独类中进行fallback处理
    @Component
    class PaymentClientFallback implements PaymentClient{
        @Override
        public String echo(Integer id) {
            return "FeignClient FallBack";
        }
    }

    @Component
    class PaymentFallbackFactory implements FallbackFactory<PaymentClient> {

        @Override
        public PaymentClient create(Throwable cause) {
            return () -> {
                System.out.println(cause.getMessage());
                return "fallbackFactory";
            };
        }
    }
}

整合DashBoard

        <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-netflix-hystrix-dashboard</artifactId>
        </dependency>

Sentinel

maven

        <dependency>
            <groupId>com.alibaba.cloud</groupId>
            <artifactId>spring-cloud-starter-alibaba-sentinel</artifactId>
        </dependency>
	<!-- sentinel-nacos--> 持久化规则
        <dependency>
            <groupId>com.alibaba.csp</groupId>
            <artifactId>sentinel-datasource-nacos</artifactId>
        </dependency>

yml

spring:
  application:
    name: nacos-provider-payment
  cloud:
    nacos:
      server-addr: localhost:8848
    sentinel: #sentinel-dashboard控制台
      transport:
        dashboard: localhost:8080
        port: 8719 # 心跳监听端口，重复会自增
      filter: # filter拦截的HttpRequest路径
        url-patterns: /**
      eager: true # 关闭懒加载
      datasource: # 持久化相关配置
        flow:
          nacos:
            serverAddr: localhost:8848
            dataId: ${spring.application.name}-flow-rule # nacos上面的dataId配置名称
            dataType: json #默认
            ruleType: flow #参考RuleType # 控制规则

指定nacos配置之后，只能通过在nacos配置中心上面添加配置才能动态持久化配置

配置属性对应具体的规则。eg : FlowRule , DegradeRule

[
    {
        "resource": "/helloworld", //资源名称
        "limitApp": "default", //流控来源
        "grade": 1, //限流阈值类型 0->线程数 1-> QPS
        "count": 2, //阈值
        "strategy": 0, //流控模式 0 直接 1 关联 2 链路
      	"refResource": "xx", //关联资源获取链路入口
        "controlBehavior": 0, //控制效果 0快速失败 1 warm up 2 排队等待
        "clusterMode": false //是否集群模式
    }
]

具体控制台的参数使用

![image-20200405183153103](/Users/hujing/Library/Application Support/typora-user-images/image-20200405183153103.png)
- 流控规则
  - QPS/线程数
  - 直接/关联(关联资源达到阈值，自己限流，比如支付扛不住了，就限流下单接口)/链路（限制链路入口的阈值，达到则限流自己）
  - 快速失败 / 预热（在配置的时间内逐渐达到阈值，防止大流量直接击垮服务）/ 排队等待（进入队列等待，超时失败）
![image-20200405183221026](/Users/hujing/Library/Application Support/typora-user-images/image-20200405183221026.png)
- 降级规则
  - RT ：超时时间
  - 异常比例 : qps>=5的时候，达到异常比例则进入熔断
  - 异常数：达到异常数进入熔断
![image-20200405183233949](/Users/hujing/Library/Application Support/typora-user-images/image-20200405183233949.png)
- 热点规则
  
  只支持基本类型和String类型
  
  可以为资源的某个参数的所有值/某写特殊值进行限流
![image-20200405183322213](/Users/hujing/Library/Application Support/typora-user-images/image-20200405183322213.png)
- 系统规则
  - 在对应系统规则达到阈值则进行限流
@SentinelResource的使用

声明Sentinel资源的注解接口，类似于HystrixCommand

   @SentinelResource(value = "hotKey",blockHandler = "hotKeyBlocked",fallback="xx"，
                    exceptionsToIgnore = "customException")
    //HotKey不支持复杂类型参数
    public String testEchoHotKey(@RequestBody HotKeyParam hotKeyParam) {
        return "blocked";
    }
    public String hotKeyBlocked(HotKeyParam hotKeyParam, BlockException exception){
        return hotKeyParam+"blocked-> "+exception.getMessage();
    }

blockHandler ：用来处理所有在sentinel-dashboard配置的参数达到阈值的兜底方法，不处理程序中的异常
fallback : 用来处理Java程序中发生异常的兜底方法
exceptionsToIgnore ：忽略的异常
核心参数
- SphU.entry() ：资源入口
- ContextUtil : 上下文
- Trace : 统计组件

4、网关

Zuul

Gateway

新一代微服务网关，代替zuul1.x

Maven依赖

        <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-gateway</artifactId>
        </dependency>

路由 : routing

断言 : predicates

在请求路由之前进行断言，若不符合则不通过路由转发，都是RoutePredicateFactory的实现类

spring:
	cloud:
    gateway:
      discovery:
        locator:
          enabled: true #开启使用服务名进行路由
      routes:
        - id: cloud-provider-payment
          #  uri: http://localhost:8001 可以写死url
          uri: lb://cloud-provider-payment #使用负载均衡服务名进行动态路由匹配
          predicates: #参考RoutePredicateFactory的实现类
            - Path=/payment/echo/{segment},/payment/**
# 支持的path映射,其中segment可以在GatewayFilter中通ServerWebExchangeUtils#getUriTemplateVariables()获取
              - Header=token,\d+ #下游服务必须要有对应的header
              - Cookie=MyCookieName,MyCookieValue #下游必须有cookie
              - Method=GET,POST #下游支持的方法
              - Query=requestParamName #下游请求参数
              - RemoteAddr=192.168.1.1/24 #规定下游的ip地址范围
              - Weight=group1,8 #访问权重，group为组名，后面为权重

过滤 : filters

主要功能是在请求之前和之后做一些增强操作，都是GatewayFilterFactory的实现类

spring:
	cloud:
    gateway:
      discovery:
        locator:
          enabled: true #开启使用服务名进行路由
      routes:
        - id: cloud-provider-payment
          #  uri: http://localhost:8001 可以写死url
          uri: lb://cloud-provider-payment #使用负载均衡服务名进行动态路由匹配
          filters:
            - AddRequestHeader=customRequestHeader,header-{segment} #这里segment可以获取上面Path断言中传递的uriTemplatesVariables
            - AddRequestParameter=customRequestParameter,xxx  #请求参数添加
            - AddResponseHeader=xxx,yyy #添加响应头
            - DedupeResponseHeader=Access-Control-Allow-Credentials Access-Control-Allow-Origin #在网关和下游都添加有CORS跨域请求头的时候，然后重复的请求头
            - name: CircuitBreaker #
              args:
                name: fallback
                fallbackUri: forward:/payment/fallback #默认fallback地址

自定义GatewayFilter

/**
 * 自定义简易网关全局过滤器
 **/
@Component
public class CustomGlobalFilter implements GlobalFilter, Ordered {
    @Override
    public Mono<Void> filter(ServerWebExchange exchange, GatewayFilterChain chain) {
        System.out.println("welcome to my spring-cloud-learning project.");
        String nameParam = exchange.getRequest().getQueryParams().getFirst("name");
        if (nameParam == null) {
            ServerHttpResponse response = exchange.getResponse();
            return response.setComplete();
        }
        return chain.filter(exchange);
    }

    @Override
    public int getOrder() {
        return 0;
    }
}

5、分布式配置中心

SpringCloud-Config

基于Git的分布式配置管理，需要自己搭建config-server端

Config-Server

pom

    <dependency>
        <groupId>org.springframework.cloud</groupId>
        <artifactId>spring-cloud-config-server</artifactId>
    </dependency>

yml

spring:
  application:
    name: config-server
  cloud: # Consul注册中心配置
    consul:
      host: localhost
      port: 8500
    config: # Config-Server的Git地址配置
      server:
        git:
          uri: https://github.com/zzhujing/cloud-config-server.git
          search-paths:
            - cloud-config-server
          username: zzhujing
          password: hujing19960825
      label: master
management:
  endpoints:
    web:
      exposure:
        include: ['bus-refresh']

配置

主程序类上添加@EnableConfigServer

Config-Client

pom

        <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-config</artifactId>
        </dependency

yml

在bootstrap.yml中指定配置

spring:
  application:
    name: cloud-provider-payment
    cloud:
      config:
        label: master #所在分支
        profile: prod #对应环境
        name: config #获取的远程配置文件名
        uri: http://localhost:8080 #Config-Server地址

加载规则

Branch/{spring.application.name}-{profile}-{properites/yml}

Branch/{spring.cloud.config.name}-{profile}-{propertiles/tml}

如何动态刷新配置？

局部手动刷新，用POST方法访问/actuator/refresh监控端点
集成spring-cloud-bus进行总线消息推送

整合`spring-cloud-bus`

maven

    <dependency> <!-- rabbitmq 的实现方式 -->
        <groupId>org.springframework.cloud</groupId>
        <artifactId>spring-cloud-starter-bus-amqp</artifactId>
    </dependency>

yml

management:
  endpoints:
    web:
      exposure:
        include: ['bus-refresh']
spring:
  rabbitmq:
  	username: guest
    password: guest
    host: localhost
    port: 5672

对config-server进行POST请求/bus-refresh即可(使用的是rabbitmq发送一个topic类型的消息，广播给各个服务的/actuator/refresh）
在需要动态刷新的配置类上面进行@RefreshScope标注

Nacos

maven

        <dependency>
            <groupId>com.alibaba.cloud</groupId>
            <artifactId>spring-cloud-starter-alibaba-nacos-config</artifactId>
        </dependency>

yml

spring:
  application:
    name: manuf-service
  cloud:
    nacos:
      config:
        ext-config: # 可以配置多个文件
          - data-id: nacos-discovery.yml # 注册到nacos上面的dataId
            refresh: true                # 是否动态刷新
          - data-id: feign-ext.yml
            refresh: true
        server-addr: 192.168.18.91:8848 # 配置中心地址
        file-extension: yml # 配置中心文件后缀格式

配置@RefreshScode即可进行动态刷新啦

6、分布式消息服务

Spring-Cloud-Stream是通过在各种消息中间件中间使用Binder进行中间桥梁搭建，让一个系统中可以通过其进行多种消息中间件的对接

rabbitMq

maven

        <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-stream-rabbit</artifactId>
        </dependency>

yml

spring:
  application:
    name: cloud-stream-provider
  cloud:
    stream:
      binders:
        defaultRabbit: #自定义Binder名称
          type: rabbit
          environment:
            spring: #rabbitmq环境配置
              rabbitmq:
                host: localhost
                port: 5672
                username: guest
                password: guest
      bindings:
        output: #自定义binding名称 提供者output / 消费者 input (也可以自定义)
          destination: customExchange #raibbitmq中Exchange名称
          content-type: application/json
          binder: defaultRabbit #定义自定义Binder名称
          group:  # 组字段用来解决重复性消费以及消息持久化问题(宕机重启之后依旧可以消费消息)

实践代码

自定义的ChannelBinding

public interface CustomerChannel {

    String OUTPUT = "my-output";   //对应spring.cloud.stream.bindings.<bindingName>
    String INPUT = "my-input";

    @Output(CustomerChannel.OUTPUT)
    MessageChannel output();

    @Input(CustomerChannel.INPUT)
    SubscribableChannel input();

}

发送消息

@EnableBinding(Source.class) //绑定提供channel名称
public class MessageProvider {

    @Autowired
    private MessageChannel output; //输出Channel，也可以直接把Source给注入进来

    public String sendMsg() {
        String msg = "Hello,Spring-Cloud-Stream";
        output.send(MessageBuilder.withPayload(msg).build());
        return null;
    }
}

接受消息

@EnableBinding(Sink.class)
public class CloudStreamConsumer {

    @Value("${server.port}")
    private String port;

    @StreamListener(Sink.INPUT) //使用监听注解监听消息
    public void received(Message<String> message) {
        System.out.println(port + " 收到消息: " + message.getPayload());
    }
}

RocketMq

kafka

7、分布式链路追踪

Sleuth + Zipkin

下载zipkin的服务端。并启动
maven

        <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-zipkin</artifactId>
        </dependency>

yml

spring:
  zipkin:
    base-url: http://localhost:9411/ # zipkin-server 地址
  sleuth:
    sampler:
      probability: 1 #采集样本所占比例 0~1

SkyWalking

8、分布式事务

seata

AT模式原理 : 两阶段提交

一阶段业务SQL和undo_log回滚日志在同一本地事务中提交，释放本地锁和连接资源
二阶段全局异步提交或根据undo_log逆向回滚

执行流程

Before image , 保存当前数据到undo_log的rollback_info字段
执行业务sql
after image , 保存执行业务SQL后的数据到undo_log

{
	"branchId": 641789253,
	"undoItems": [{
		"afterImage": {
			"rows": [{
				"fields": [{
					"name": "id",
					"type": 4,
					"value": 1
				}, {
					"name": "name",
					"type": 12,
					"value": "GTS"
				}, {
					"name": "since",
					"type": 12,
					"value": "2014"
				}]
			}],
			"tableName": "product"
		},
		"beforeImage": {
			"rows": [{
				"fields": [{
					"name": "id",
					"type": 4,
					"value": 1
				}, {
					"name": "name",
					"type": 12,
					"value": "TXC"
				}, {
					"name": "since",
					"type": 12,
					"value": "2014"
				}]
			}],
			"tableName": "product"
		},
		"sqlType": "UPDATE"
	}],
	"xid": "xid:xxx"
}

提交本地事务前获取全局锁，注册分支。
提交SQL和undo_log然后上报给TC
二阶段异步提交，删除undo_log记录
二阶段回滚，根据全局id和分支id获取undo_log，逆向生成sql进行回滚，若after image和当前数据库不一样，则发生脏写(幻读)。转人工处理！

搭建seata-server

下载seata-server

修改conf/registry.conf，使用nacos作为注册和配置中心，seata支持file,nacos,eureka,redis,zk,consul,sofa...

使用官网提供的nacos-config.sh将config.txt中的seata配置写入到nacos配置中心

transport.type=TCP
transport.server=NIO
transport.heartbeat=true
transport.enable-client-batch-send-request=false
transport.thread-factory.boss-thread-prefix=NettyBoss
transport.thread-factory.worker-thread-prefix=NettyServerNIOWorker
transport.thread-factory.server-executor-thread-prefix=NettyServerBizHandler
transport.thread-factory.share-boss-worker=false
transport.thread-factory.client-selector-thread-prefix=NettyClientSelector
transport.thread-factory.client-selector-thread-size=1
transport.thread-factory.client-worker-thread-prefix=NettyClientWorkerThread
transport.thread-factory.boss-thread-size=1
transport.thread-factory.worker-thread-size=8
transport.shutdown.wait=3
#这里很重要，配置和spring.cloud.alibaba.seata.tx_service_group相同，可配置多个
service.vgroup_mapping.test_group=default          
service.default.grouplist=127.0.0.1:8091
service.enableDegrade=false
service.disableGlobalTransaction=false
client.rm.async.commit.buffer.limit=10000
client.rm.lock.retry.internal=10
client.rm.lock.retry.times=30
client.rm.report.retry.count=5
client.rm.lock.retry.policy.branch-rollback-on-conflict=true
client.rm.table.meta.check.enable=false
client.rm.report.success.enable=true
client.tm.commit.retry.count=5
client.tm.rollback.retry.count=5
#事务存储模式
store.mode=db 																				
store.file.dir=file_store/data
store.file.max-branch-session-size=16384
store.file.max-global-session-size=512
store.file.file-write-buffer-cache-size=16384
store.file.flush-disk-mode=async
store.file.session.reload.read_size=100
#数据库相关配置
store.db.datasource=dbcp  													
store.db.db-type=mysql
store.db.driver-class-name=com.mysql.cj.jdbc.Driver
store.db.url=jdbc:mysql://127.0.0.1:3307/seata?useUnicode=true
store.db.user=hujing
store.db.password=hujing19960825
store.db.min-conn=1
store.db.max-conn=3
store.db.global.table=global_table
store.db.branch.table=branch_table
store.db.query-limit=100
store.db.lock-table=lock_table
server.recovery.committing-retry-period=1000
server.recovery.asyn-committing-retry-period=1000
server.recovery.rollbacking-retry-period=1000
server.recovery.timeout-retry-period=1000
server.max.commit.retry.timeout=-1
server.max.rollback.retry.timeout=-1
client.undo.data.validation=true
client.undo.log.serialization=jackson
server.undo.log.save.days=7
server.undo.log.delete.period=86400000
client.undo.log.table=undo_log
client.log.exceptionRate=100
transport.serialization=seata
transport.compressor=none
metrics.enabled=false
metrics.registry-type=compact
metrics.exporter-list=prometheus
metrics.exporter-prometheus-port=9898
#是否支持自动数据库代理
client.support.spring.datasource.autoproxy=false

sh nacos-config.sh 将配置同步到配置中心，完成后再nacos上面能看到配置即可
这里使用nacos做为seata的配置中心，所以file.conf文件我们可以不用操作，否则需要修改**

service {
#事务组名，和spring.cloud.alibaba.seata.tx-service-group相同
  vgroup_mapping.my_test_tx_group = "default" 
  #only support when registry.type=file, please don't set multiple addresses
  default.grouplist = "127.0.0.1:8091"
  #disable seata
  disableGlobalTransaction = false
}
## transaction log store, only used in seata-server
store {
  ## store mode: file、db
  mode = "db" #事务日志存储模式,使用db

  ## file store property
  file {
    ## store location dir
    dir = "sessionStore"
  }
  db {
    datasource = "dbcp" 
    db-type = "mysql"
    #可以改成mysql-8，启动server的时候需要添加mysql8的jar包到lib目录
    driver-class-name = "com.mysql.jdbc.Driver" 
    url = "jdbc:mysql://127.0.0.1:3306/seata" #seata-server依赖的数据库的位置
    user = "mysql"
    password = "mysql"
  }
}

启动seata-server即可，如果nacos上面有serverAddr的服务实例

Client

maven

注意版本，版本对不上会有很多的问题发生

        <dependency>
            <groupId>com.alibaba.cloud</groupId>
            <artifactId>spring-cloud-starter-alibaba-seata</artifactId>
        </dependency>

yml

spring:
  application:
    name: seata-storage
  cloud:
    nacos:
      server-addr: localhost:8848
    alibaba:
      seata:
        tx-service-group: test_group

在resource类路径下添加registry.conf文件即可

registry {
  # file 、nacos 、eureka、redis、zk、consul、etcd3、sofa
  type = "nacos"
  nacos {
    serverAddr = "localhost"
    namespace = ""
    cluster = "default"
  }
}
config {
  # file、nacos 、apollo、zk、consul、etcd3
  type = "nacos"
  nacos {
    serverAddr = "localhost"
    namespace = ""
    group = "SEATA_GROUP"
  }
}

在TM业务逻辑加上@GlobalTransactional

    /**
     * 下单
     *
     * @param order 订单相关信息
     */
    @Override
    @GlobalTransactional(rollbackFor = Exception.class,name="test_gloabl_tx")
    public void placeAnOrder(Order order) {
        //扣除库存
        storageClient.reduceStorage(order.getProductId(), order.getCount());
        //扣余额
        accountClient.reduceAccount(order.getUserId(), 10);
        //创建订单
        createOrder(order);
    }

Name		Name	Last commit message	Last commit date
cloud-common-api		cloud-common-api
cloud-config-server		cloud-config-server
cloud-consumer-order80		cloud-consumer-order80
cloud-gateway		cloud-gateway
cloud-provider-payment8001		cloud-provider-payment8001
cloud-provider-payment8004		cloud-provider-payment8004
cloud-registry-server7001		cloud-registry-server7001
cloud-registry-server7002		cloud-registry-server7002
cloud-seata-account		cloud-seata-account
cloud-seata-common		cloud-seata-common
cloud-seata-order		cloud-seata-order
cloud-seata-storage		cloud-seata-storage
cloud-stream-consumer8082		cloud-stream-consumer8082
cloud-stream-provider8081		cloud-stream-provider8081
cloudalibaba-consumer-order90		cloudalibaba-consumer-order90
cloudalibaba-provider-payment9001		cloudalibaba-provider-payment9001
common-datasource-configuration		common-datasource-configuration
.gitignore		.gitignore
README.md		README.md
pom.xml		pom.xml

Latest commit

History

Folders and files

Repository files navigation

SpringCloud

CAP

1、服务注册与发现

Eureka

server

环境搭建

client

Zookeeper

server

client

Consul

Server

Client

Nacos

Server

使用官方的nacos包进行集群搭建

基于docker搭建Nacos-Server

Client

2、远程调用

Ribbon+RestTemplate

Feign+Ribbon

3、服务容错、熔断、限流

Hystrix

容错支持@HystrixCommand

熔断，断路器模式

解决太多熔断方法膨胀代码？

解决fallback方法和业务耦合在一起

整合DashBoard

Sentinel

4、网关

Zuul

Gateway

Maven依赖

路由 : routing

断言 : predicates

过滤 : filters

自定义GatewayFilter

5、分布式配置中心

SpringCloud-Config

Config-Server

Config-Client

如何动态刷新配置？

整合spring-cloud-bus

Nacos

6、分布式消息服务

rabbitMq

自定义的ChannelBinding

发送消息

接受消息

RocketMq

kafka

7、分布式链路追踪

Sleuth + Zipkin

SkyWalking

8、分布式事务

seata

AT模式原理 : 两阶段提交

搭建seata-server

Client

About

Resources

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages

容错支持`@HystrixCommand`

整合`spring-cloud-bus`