公众号「架构成长指南」,专注于生产实践、云原生、分布式系统、大数据技术分享。

目的

Spring Cloud 线上微服务实例都是2个起步,如果出问题后,在没有ELK等日志分析平台,如何确定调用到了目标服务的那个实例,以此来排查问题

效果

可以看到服务有几个实例是上线,并且最终调用了那个实例

考虑到Spring Cloud在版本升级中使用了两种负载均衡实现,RobinLoadBalancer,下面我们提供两种实现方案

Robin实现方案1. 技术栈

  • Spring Cloud: Hoxton.SR6
  • Spring Boot: 2.3.1.RELEASE
  • Spring-Cloud-Openfeign: 2.2.3.RELEASE

2. 继承RoundRobinRule,并重写choose方法

/** * 因为调用目标机器的时候,如果目标机器本身假死或者调用目标不通无法数据返回,那么feign无法打印目标机器。这种场景下我们需要在调用失败(目标机器没有返回)的时候也能把目标机器的ip打印出来,这种场景需要我们切入feign选择机器的逻辑,注入我们自己的调度策略(默认是roundrobin),在里面打印选择的机器即可。*/@Slf4jpublic class FeignRule extends RoundRobinRule {    @Override    public Server choose(Object key) {        Server server = super.choose(key);        if (Objects.isNull(server)) {            log.info("server is null");            return null;        }        log.info("feign rule ---> serverName:{}, choose key:{}, final server ip:{}", server.getMetaInfo().getAppName(), key, server.getHostPort());        return server;    }    @Override    public Server choose(ILoadBalancer lb, Object key) {        Server chooseServer = super.choose(lb, key);        List reachableServers = lb.getReachableServers();        List allServers = lb.getAllServers();        int upCount = reachableServers.size();        int serverCount = allServers.size();        log.info("serverName:{} upCount:{}, serverCount:{}", Objects.nonNull(chooseServer) ? chooseServer.getMetaInfo().getAppName() : "", upCount, serverCount);        for (Server server : allServers) {            if (server instanceof DiscoveryEnabledServer) {                DiscoveryEnabledServer dServer = (DiscoveryEnabledServer) server;                InstanceInfo instanceInfo = dServer.getInstanceInfo();                if (instanceInfo != null) {                    InstanceInfo.InstanceStatus status = instanceInfo.getStatus();                    if (status != null) {                        log.info("serverName:{} server:{}, status:{}", server.getMetaInfo().getAppName(), server.getHostPort(), status);                    }                }            }        }        return chooseServer;    }}

3.修改RibbonClients配置

import org.springframework.cloud.netflix.ribbon.RibbonClients;import org.springframework.context.annotation.Configuration;/** * @description:feign 配置 */@Configuration@RibbonClients(defaultConfiguration = {FeignRule.class})public class FeignConfig {}

LoadBalancer实现方案1. 技术栈

  • Spring Cloud: 2021.0.4
  • Spring Boot: 2.7.17
  • Spring-Cloud-Openfeign: 3.1.4

2. 继承ReactorServiceInstanceLoadBalancer,并实现相关方法

@Slf4jpublic class CustomRoundRobinLoadBalancer implements ReactorServiceInstanceLoadBalancer {    final AtomicInteger position;    final String serviceId;    ObjectProvider serviceInstanceListSupplierProvider;    public CustomRoundRobinLoadBalancer(ObjectProvider serviceInstanceListSupplierProvider, String serviceId) {        this(serviceInstanceListSupplierProvider, serviceId, (new Random()).nextInt(1000));    }    public CustomRoundRobinLoadBalancer(ObjectProvider serviceInstanceListSupplierProvider, String serviceId, int seedPosition) {        this.serviceId = serviceId;        this.serviceInstanceListSupplierProvider = serviceInstanceListSupplierProvider;        this.position = new AtomicInteger(seedPosition);    }    public Mono<Response> choose(Request request) {        ServiceInstanceListSupplier supplier = this.serviceInstanceListSupplierProvider.getIfAvailable(NoopServiceInstanceListSupplier::new);        return supplier.get(request).next().map((serviceInstances) -> {            return this.processInstanceResponse(supplier, serviceInstances);        });    }    private Response processInstanceResponse(ServiceInstanceListSupplier supplier, List serviceInstances) {        Response serviceInstanceResponse = this.getInstanceResponse(serviceInstances);        if (supplier instanceof SelectedInstanceCallback && serviceInstanceResponse.hasServer()) {            ((SelectedInstanceCallback)supplier).selectedServiceInstance((ServiceInstance)serviceInstanceResponse.getServer());        }        return serviceInstanceResponse;    }    private Response getInstanceResponse(List instances) {        if (instances.isEmpty()) {            if (log.isWarnEnabled()) {                log.warn("No servers available for service: " + this.serviceId);            }            return new EmptyResponse();        } else {            int pos = this.position.incrementAndGet() & Integer.MAX_VALUE;            ServiceInstance instance = instances.get(pos % instances.size());            log.info("serverName:{} upCount:{}",instance.getServiceId(),instances.size());            log.info("feign rule ---> serverName:{}, final server ip:{}:{}", instance.getServiceId(), instance.getHost(),instance.getPort());            return new DefaultResponse(instance);        }    }}

2.修改LoadBalancerClients配置

@Configuration@LoadBalancerClients(defaultConfiguration = CustomLoadBalancerConfiguration.class)public class CustomLoadBalancerConfig {}@Configurationclass CustomLoadBalancerConfiguration {    /**     * 参考默认实现     * @see org.springframework.cloud.loadbalancer.annotation.LoadBalancerClientConfiguration#reactorServiceInstanceLoadBalancer     * @return     */    @Bean    public ReactorLoadBalancer reactorServiceInstanceLoadBalancer(Environment environment, LoadBalancerClientFactory loadBalancerClientFactory) {        String name = environment.getProperty(LoadBalancerClientFactory.PROPERTY_NAME);        return new CustomRoundRobinLoadBalancer(loadBalancerClientFactory.getLazyProvider(name, ServiceInstanceListSupplier.class), name);    }}

以上两部完成大功告成!

源码下载:

https://github.com/dongweizhao/spring-cloud-example/tree/SR6-OpenFeign
https://github.com/dongweizhao/spring-cloud-example/tree/EurekaOpenFeign