Spring-Cloud-Ribbon学习笔记（二）：自定义负载均衡规则

Ribbon自定义负载均衡策略有两种方式，一是JavaConfig，一是通过配置文件（yml或properties文件）。

需求

假设我有包含A和B服务在内的多个微服务，它们均注册在一个Eureka上，信息如下：

我希望当访问服务A时候，2个服务（端口分别是8087和8081）每两次一换，比如访问两次8087，再访问两次8081，如此反复。

当访问服务B时，与A类似，不过是3次一换。

当访问其他服务时，采用随机规则，即RandomRule，而不是默认策略^[1]

JavaConfig

使用这种方式，总共分3步（以服务A的规则为例）：

• 新建针对服务A的负载均衡规则类SvcARule，实现抽象类AbstractLoadBalancerRule并重写方法，在类上只需要添加注解@Primary^[2]，重写逻辑须要根据实际需求来定，详见下面的代码
• 新建配置类LBConfig，添加类型为IRule的Bean，此处我们选RandomRule
• 在配置类LBConfig上,通过注解@RibbonClients（只有一种自定义规则，则使用@RibbonClient），添加不适用公共规则的其他自定义规则

实现

SvcARule

import com.netflix.client.config.IClientConfig;
import com.netflix.loadbalancer.AbstractLoadBalancerRule;
import com.netflix.loadbalancer.ILoadBalancer;
import com.netflix.loadbalancer.Server;
import org.springframework.context.annotation.Primary;

import java.util.List;
import java.util.concurrent.atomic.AtomicInteger;

@Primary
public class SvcARule extends AbstractLoadBalancerRule {

    private AtomicInteger total = new AtomicInteger();
    private AtomicInteger index = new AtomicInteger();

    public Server choose(ILoadBalancer lb, Object key) {
        if (lb == null) {
            return null;
        }
        Server server = null;

        while (server == null) {
            if (Thread.interrupted()) {
                return null;
            }
            List<Server> upList = lb.getReachableServers();
            List<Server> allList = lb.getAllServers();

            int serverCount = allList.size();
            if (serverCount == 0) {
                return null;
            }
            if (total.get() < 2) {
                server = upList.get(index.get());
                total.incrementAndGet();
            } else {
                total.set(0);
                index.set(index.incrementAndGet() % upList.size());
            }

            if (server == null) {
                Thread.yield();
                continue;
            }

            if (server.isAlive()) {
                return (server);
            }

            server = null;
            Thread.yield();
        }
        return server;

    }

    @Override
    public Server choose(Object key) {
        return choose(getLoadBalancer(), key);
    }

    @Override
    public void initWithNiwsConfig(IClientConfig clientConfig) {
        // TODO Auto-generated method stub
    }
}

SvcBRule

略，只需要把if (total.get() < 2)改成if (total.get() < 3)即可

LBConfig

import com.netflix.loadbalancer.IRule;
import com.netflix.loadbalancer.RandomRule;
import com.yan.ribbon.rule.SvcARule;
import com.yan.ribbon.rule.SvcBRule;
import org.springframework.cloud.netflix.ribbon.RibbonClient;
import org.springframework.cloud.netflix.ribbon.RibbonClients;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;

@Configuration
@RibbonClients(
        value = {
                @RibbonClient(name = "SVCA", configuration = SvcARule.class),
                @RibbonClient(name = "SVCB", configuration = SvcBRule.class)
        },
        defaultConfiguration = LBConfig.class)
public class LBConfig {
    @Bean
    public IRule commonRule() {
        return new RandomRule();
    }
}

测试

通过postman多次访问服务A、B和C，得到日志如下：

2019-03-19 20:22:09.511  INFO 13372 --- [nio-8004-exec-4] o.a.c.c.C.[Tomcat].[localhost].[/]       : Initializing Spring DispatcherServlet 'dispatcherServlet'
2019-03-19 20:22:09.511  INFO 13372 --- [nio-8004-exec-4] o.s.web.servlet.DispatcherServlet        : Initializing Servlet 'dispatcherServlet'
2019-03-19 20:22:09.519  INFO 13372 --- [nio-8004-exec-4] o.s.web.servlet.DispatcherServlet        : Completed initialization in 8 ms
2019-03-19 20:22:09.693  INFO 13372 --- [nio-8004-exec-4] c.n.u.concurrent.ShutdownEnabledTimer    : Shutdown hook installed for: NFLoadBalancer-PingTimer-SVCA
2019-03-19 20:22:09.708  INFO 13372 --- [nio-8004-exec-4] c.netflix.loadbalancer.BaseLoadBalancer  : Client: SVCA instantiated a LoadBalancer: DynamicServerListLoadBalancer:{NFLoadBalancer:name=SVCA,current list of Servers=[],Load balancer stats=Zone stats: {},Server stats: []}ServerList:null
2019-03-19 20:22:09.712  INFO 13372 --- [nio-8004-exec-4] c.n.l.DynamicServerListLoadBalancer      : Using serverListUpdater PollingServerListUpdater
2019-03-19 20:22:09.729  INFO 13372 --- [nio-8004-exec-4] c.n.l.DynamicServerListLoadBalancer      : DynamicServerListLoadBalancer for client SVCA initialized: DynamicServerListLoadBalancer:{NFLoadBalancer:name=SVCA,current list of Servers=[YanWei:8087, YanWei:8081],Load balancer stats=Zone stats: {defaultzone=[Zone:defaultzone;	Instance count:2;	Active connections count: 0;	Circuit breaker tripped count: 0;	Active connections per server: 0.0;]
},Server stats: [[Server:YanWei:8087;	Zone:defaultZone;	Total Requests:0;	Successive connection failure:0;	Total blackout seconds:0;	Last connection made:Thu Jan 01 08:00:00 CST 1970;	First connection made: Thu Jan 01 08:00:00 CST 1970;	Active Connections:0;	total failure count in last (1000) msecs:0;	average resp time:0.0;	90 percentile resp time:0.0;	95 percentile resp time:0.0;	min resp time:0.0;	max resp time:0.0;	stddev resp time:0.0]
, [Server:YanWei:8081;	Zone:defaultZone;	Total Requests:0;	Successive connection failure:0;	Total blackout seconds:0;	Last connection made:Thu Jan 01 08:00:00 CST 1970;	First connection made: Thu Jan 01 08:00:00 CST 1970;	Active Connections:0;	total failure count in last (1000) msecs:0;	average resp time:0.0;	90 percentile resp time:0.0;	95 percentile resp time:0.0;	min resp time:0.0;	max resp time:0.0;	stddev resp time:0.0]
]}ServerList:org.springframework.cloud.netflix.ribbon.eureka.DomainExtractingServerList@7f0246c6
"svcA:8087===>\nname:xixi\ntype:haha"
"svcA:8087===>\nname:xixi\ntype:haha"
"svcA:8081===>\nname:xixi\ntype:haha"
"svcA:8081===>\nname:xixi\ntype:haha"
"svcA:8087===>\nname:xixi\ntype:haha"
"svcA:8087===>\nname:xixi\ntype:haha"
"svcA:8081===>\nname:xixi\ntype:haha"
"svcA:8081===>\nname:xixi\ntype:haha"
"svcA:8087===>\nname:xixi\ntype:haha"
"svcA:8087===>\nname:xixi\ntype:haha"
"svcA:8081===>\nname:xixi\ntype:haha"
"svcA:8081===>\nname:xixi\ntype:haha"
2019-03-19 20:22:37.173  INFO 13372 --- [nio-8004-exec-5] c.n.u.concurrent.ShutdownEnabledTimer    : Shutdown hook installed for: NFLoadBalancer-PingTimer-SVCB
2019-03-19 20:22:37.174  INFO 13372 --- [nio-8004-exec-5] c.netflix.loadbalancer.BaseLoadBalancer  : Client: SVCB instantiated a LoadBalancer: DynamicServerListLoadBalancer:{NFLoadBalancer:name=SVCB,current list of Servers=[],Load balancer stats=Zone stats: {},Server stats: []}ServerList:null
2019-03-19 20:22:37.177  INFO 13372 --- [nio-8004-exec-5] c.n.l.DynamicServerListLoadBalancer      : Using serverListUpdater PollingServerListUpdater
2019-03-19 20:22:37.178  INFO 13372 --- [nio-8004-exec-5] c.n.l.DynamicServerListLoadBalancer      : DynamicServerListLoadBalancer for client SVCB initialized: DynamicServerListLoadBalancer:{NFLoadBalancer:name=SVCB,current list of Servers=[YanWei:8086, YanWei:8082],Load balancer stats=Zone stats: {defaultzone=[Zone:defaultzone;	Instance count:2;	Active connections count: 0;	Circuit breaker tripped count: 0;	Active connections per server: 0.0;]
},Server stats: [[Server:YanWei:8082;	Zone:defaultZone;	Total Requests:0;	Successive connection failure:0;	Total blackout seconds:0;	Last connection made:Thu Jan 01 08:00:00 CST 1970;	First connection made: Thu Jan 01 08:00:00 CST 1970;	Active Connections:0;	total failure count in last (1000) msecs:0;	average resp time:0.0;	90 percentile resp time:0.0;	95 percentile resp time:0.0;	min resp time:0.0;	max resp time:0.0;	stddev resp time:0.0]
, [Server:YanWei:8086;	Zone:defaultZone;	Total Requests:0;	Successive connection failure:0;	Total blackout seconds:0;	Last connection made:Thu Jan 01 08:00:00 CST 1970;	First connection made: Thu Jan 01 08:00:00 CST 1970;	Active Connections:0;	total failure count in last (1000) msecs:0;	average resp time:0.0;	90 percentile resp time:0.0;	95 percentile resp time:0.0;	min resp time:0.0;	max resp time:0.0;	stddev resp time:0.0]
]}ServerList:org.springframework.cloud.netflix.ribbon.eureka.DomainExtractingServerList@485ed210
"svcB:8086===>\nname:xixi\ntype:haha"
"svcB:8086===>\nname:xixi\ntype:haha"
"svcB:8086===>\nname:xixi\ntype:haha"
"svcB:8082===>\nname:xixi\ntype:haha"
"svcB:8082===>\nname:xixi\ntype:haha"
"svcB:8082===>\nname:xixi\ntype:haha"
"svcB:8086===>\nname:xixi\ntype:haha"
"svcB:8086===>\nname:xixi\ntype:haha"
"svcB:8086===>\nname:xixi\ntype:haha"
"svcB:8082===>\nname:xixi\ntype:haha"
"svcB:8082===>\nname:xixi\ntype:haha"
"svcB:8082===>\nname:xixi\ntype:haha"
"svcB:8086===>\nname:xixi\ntype:haha"
"svcB:8086===>\nname:xixi\ntype:haha"
2019-03-19 20:23:01.319  INFO 13372 --- [nio-8004-exec-8] c.n.u.concurrent.ShutdownEnabledTimer    : Shutdown hook installed for: NFLoadBalancer-PingTimer-SVCC
2019-03-19 20:23:01.320  INFO 13372 --- [nio-8004-exec-8] c.netflix.loadbalancer.BaseLoadBalancer  : Client: SVCC instantiated a LoadBalancer: DynamicServerListLoadBalancer:{NFLoadBalancer:name=SVCC,current list of Servers=[],Load balancer stats=Zone stats: {},Server stats: []}ServerList:null
2019-03-19 20:23:01.320  INFO 13372 --- [nio-8004-exec-8] c.n.l.DynamicServerListLoadBalancer      : Using serverListUpdater PollingServerListUpdater
2019-03-19 20:23:01.322  INFO 13372 --- [nio-8004-exec-8] c.n.l.DynamicServerListLoadBalancer      : DynamicServerListLoadBalancer for client SVCC initialized: DynamicServerListLoadBalancer:{NFLoadBalancer:name=SVCC,current list of Servers=[YanWei:8083, YanWei:8085],Load balancer stats=Zone stats: {defaultzone=[Zone:defaultzone;	Instance count:2;	Active connections count: 0;	Circuit breaker tripped count: 0;	Active connections per server: 0.0;]
},Server stats: [[Server:YanWei:8083;	Zone:defaultZone;	Total Requests:0;	Successive connection failure:0;	Total blackout seconds:0;	Last connection made:Thu Jan 01 08:00:00 CST 1970;	First connection made: Thu Jan 01 08:00:00 CST 1970;	Active Connections:0;	total failure count in last (1000) msecs:0;	average resp time:0.0;	90 percentile resp time:0.0;	95 percentile resp time:0.0;	min resp time:0.0;	max resp time:0.0;	stddev resp time:0.0]
, [Server:YanWei:8085;	Zone:defaultZone;	Total Requests:0;	Successive connection failure:0;	Total blackout seconds:0;	Last connection made:Thu Jan 01 08:00:00 CST 1970;	First connection made: Thu Jan 01 08:00:00 CST 1970;	Active Connections:0;	total failure count in last (1000) msecs:0;	average resp time:0.0;	90 percentile resp time:0.0;	95 percentile resp time:0.0;	min resp time:0.0;	max resp time:0.0;	stddev resp time:0.0]
]}ServerList:org.springframework.cloud.netflix.ribbon.eureka.DomainExtractingServerList@77b35c9
"svcC:8085===>\nname:xixi\ntype:haha"
"svcC:8083===>\nname:xixi\ntype:haha"
"svcC:8085===>\nname:xixi\ntype:haha"
"svcC:8085===>\nname:xixi\ntype:haha"
"svcC:8083===>\nname:xixi\ntype:haha"
"svcC:8085===>\nname:xixi\ntype:haha"
"svcC:8085===>\nname:xixi\ntype:haha"

配置文件方式

配置文件方式须要遵守以下格式：

<serviceName>.ribbon.<key>=<value>

几个变量解释如下：

serviceName就是服务名

key是指Ribbon为实现不同领域（我们这里只说IRule，其实还有IPing等）所指定的关联类名，这里我需要自定义IRule，其关联类名，也就是所谓的key，是NFLoadBalancerRuleClassName^[3]

value是我所要指定的自定义规则，比如，SvcB的负载均衡策略SvcBRule

因此，在配置文件中添加如下配置:

svcB:
  ribbon:
    NFLoadBalancerRuleClassName: com.yan.ribbon.rule.SvcBRule

重新启动Ribbon服务后，发现此配置并没有生效，原因有两个：

一是我的服务名字写的是svcB，理由是我的配置spring.appliaction.name=svcB，但是使用Ribbon调用时使用了http://SVCB/...，SVCB是大写的，所以没有得到期望的效果^[4]

二是我已经有了用来替换默认规则的LBConfig，使得IRule的默认配置没有生效，因为查看源码org.springframework.cloud.netflix.ribbon.RibbonClientConfiguration中对于IRule的定义如下：

	@Bean
	@ConditionalOnMissingBean
	public IRule ribbonRule(IClientConfig config) {
		if (this.propertiesFactory.isSet(IRule.class, name)) {
			return this.propertiesFactory.get(IRule.class, config, name);
		}
		ZoneAvoidanceRule rule = new ZoneAvoidanceRule();
		rule.initWithNiwsConfig(config);
		return rule;
	}

由于我自定义的commonRule已经存在了，所以此处不会触发，更不会走到判断那一步，就是说，我的上面的关于SvcB的自定义规则的配置，是没有什么地方去读的，因此，为了方便演示，直接将LBConfig全部注掉，再次重启后，把访问Uri中的大写改成一致的svcB，发现对于服务B的配置生效了，但是没有指定的服务，全部都采用了默认的策略ZoneAvoidanceRule。

总结

配置文件方式，比较麻烦，且不能-没必要也懒得研究如何-修改默认规则，而JavaConfig可以完美替代，因此我选JavaConfig。

进阶

关于Ribbon中对于IRule的实现，总共有如下几种，其简略说明也附注如下：

• RoundRobinRule: 轮询
• RandomRule：随机
• AvailabilityFilteringRule：会先过滤掉由于多次访问故障而处于断路器跳闸状态的服务，还有并发的连接数超过阈值的服务，然后对剩余的服务列表按照轮询策略进行访问
• WeightedResponseTimeRule：根据平均响应时间计算所有服务的权重，响应时间越快服务权重越大，被选中的概率越高。刚启动时如果统计信息不足，则使用RoundRobinRule策略，等统计信息足够，会切换到WeightedResponseTimeRule
• RetryRule: 先按照RoundRobinRule的策略获取服务，如果获取服务失败，则在指定时间内会进行重试，获取可用服务
• BestAvailableRule: 会先过滤掉由于多次访问故障而处于断路器跳闸状态的服务，然后选择一个并发量最小的服务
• ZoneAvoidanceRule: 默认规则，符合判断server所在区域的性能和server的可用性选择服务器

---

1. 对应类为ZoneAvoidanceRule，直译过来是区域回避规则，我也不知道为啥这样叫:D ↩︎

2. 如果不加此注解，会报错，内容是找类型为IRule的Bean，但是找到了俩，一个commonRule，一个svcARule。如果你不嫌麻烦，可以选择网上流行的较为复杂的形式，比如springcloud-04-自定义ribbon的配置方式 ↩︎

3. 可以在类org.springframework.cloud.netflix.ribbon.PropertiesFactory中查看 ↩︎

4. 这个Eureka也要背锅，因为使用服务名来进行服务间调用，是被大小写是被兼容过的，导致我这边疏忽大意了，还是要细心的好。 ↩︎