java版spring cloud+spring boot 社交电子商务平台-Ribbon

社交电商平台源码请加企鹅求求:一零三八七七四六二六。 一. Spring集成下的Ribbon工作结构先贴一张总览图,说明一下Spring如何集成Ribbon的,如下所示:Spring Cloud集成模式下的Ribbon有以下...

社交电商平台源码请加企鹅求求:一零三八七七四六二六。 一. Spring集成下的Ribbon工作结构


先贴一张总览图,说明一下Spring如何集成Ribbon的,如下所示:

Ribbon工作结构.png


Spring Cloud集成模式下的Ribbon有以下几个特征:


1.Ribbon 服务配置方式


每一个服务配置都有一个Spring ApplicationContext上下文,用于加载各自服务的实例。


2. 和Feign的集成模式


在使用Feign作为客户端时,最终请求会转发成 http://<服务名称>/<relative-path-to-service>的格式,通过LoadBalancerFeignClient, 提取出服务标识<服务名称>,然后根据服务名称在上下文中查找对应服务的负载均衡器FeignLoadBalancer,负载均衡器负责根据既有的服务实例的统计信息,挑选出最合适的服务实例。


二、Spring Cloud模式下和Feign的集成实现方式


和Feign结合的场景下,Feign的调用会被包装成调用请求LoadBalancerCommand,然后底层通过Rxjava基于事件的编码风格,发送请求;Spring Cloud框架通过 Feigin 请求的URL,提取出服务名称,然后在上下文中找到对应服务的的负载均衡器实现FeignLoadBalancer,然后通过负载均衡器中挑选一个合适的Server实例,然后将调用请求转发到该Server实例上,完成调用,在此过程中,记录对应Server实例的调用统计信息。

/**
     * Create an {@link Observable} that once subscribed execute network call asynchronously with a server chosen by load balancer.
     * If there are any errors that are indicated as retriable by the {@link RetryHandler}, they will be consumed internally by the
     * function and will not be observed by the {@link Observer} subscribed to the returned {@link Observable}. If number of retries has
     * exceeds the maximal allowed, a final error will be emitted by the returned {@link Observable}. Otherwise, the first successful
     * result during execution and retries will be emitted.
     */
    public Observable<T> submit(final ServerOperation<T> operation) {
        final ExecutionInfoContext context = new ExecutionInfoContext();
        
        if (listenerInvoker != null) {
            try {
                listenerInvoker.onExecutionStart();
            } catch (AbortExecutionException e) {
                return Observable.error(e);
            }
        }
        
        // 同一Server最大尝试次数
        final int maxRetrysSame = retryHandler.getMaxRetriesOnSameServer();
        //下一Server最大尝试次数
        final int maxRetrysNext = retryHandler.getMaxRetriesOnNextServer();

        // Use the load balancer
        // 使用负载均衡器,挑选出合适的Server,然后执行Server请求,将请求的数据和行为整合到ServerStats中
        Observable<T> o = 
                (server == null ? selectServer() : Observable.just(server))
                .concatMap(new Func1<Server, Observable<T>>() {
                    @Override
                    // Called for each server being selected
                    public Observable<T> call(Server server) {
                        // 获取Server的统计值
                        context.setServer(server);
                        final ServerStats stats = loadBalancerContext.getServerStats(server);
                        
                        // Called for each attempt and retry 服务调用
                        Observable<T> o = Observable
                                .just(server)
                                .concatMap(new Func1<Server, Observable<T>>() {
                                    @Override
                                    public Observable<T> call(final Server server) {
                                        context.incAttemptCount();//重试计数
                                        loadBalancerContext.noteOpenConnection(stats);//链接统计
                                        
                                        if (listenerInvoker != null) {
                                            try {
                                                listenerInvoker.onStartWithServer(context.toExecutionInfo());
                                            } catch (AbortExecutionException e) {
                                                return Observable.error(e);
                                            }
                                        }
                                        //执行监控器,记录执行时间
                                        final Stopwatch tracer = loadBalancerContext.getExecuteTracer().start();
                                        //找到合适的server后,开始执行请求
                                        //底层调用有结果后,做消息处理
                                        return operation.call(server).doOnEach(new Observer<T>() {
                                            private T entity;
                                            @Override
                                            public void onCompleted() {
                                                recordStats(tracer, stats, entity, null);
                                                // 记录统计信息
                                            }

                                            @Override
                                            public void onError(Throwable e) {
                                                recordStats(tracer, stats, null, e);//记录异常信息
                                                logger.debug("Got error {} when executed on server {}", e, server);
                                                if (listenerInvoker != null) {
                                                    listenerInvoker.onExceptionWithServer(e, context.toExecutionInfo());
                                                }
                                            }

                                            @Override
                                            public void onNext(T entity) {
                                                this.entity = entity;//返回结果值
                                                if (listenerInvoker != null) {
                                                    listenerInvoker.onExecutionSuccess(entity, context.toExecutionInfo());
                                                }
                                            }                            
                                            
                                            private void recordStats(Stopwatch tracer, ServerStats stats, Object entity, Throwable exception) {
                                                tracer.stop();//结束计时
                                                //标记请求结束,更新统计信息
                                                loadBalancerContext.noteRequestCompletion(stats, entity, exception, tracer.getDuration(TimeUnit.MILLISECONDS), retryHandler);
                                            }
                                        });
                                    }
                                });
                        //如果失败,根据重试策略触发重试逻辑
                        // 使用observable 做重试逻辑,根据predicate 做逻辑判断,这里做
                        if (maxRetrysSame > 0) 
                            o = o.retry(retryPolicy(maxRetrysSame, true));
                        return o;
                    }
                });
         // next请求处理,基于重试器操作   
        if (maxRetrysNext > 0 && server == null) 
            o = o.retry(retryPolicy(maxRetrysNext, false));
        
        return o.onErrorResumeNext(new Func1<Throwable, Observable<T>>() {
            @Override
            public Observable<T> call(Throwable e) {
                if (context.getAttemptCount() > 0) {
                    if (maxRetrysNext > 0 && context.getServerAttemptCount() == (maxRetrysNext + 1)) {
                        e = new ClientException(ClientException.ErrorType.NUMBEROF_RETRIES_NEXTSERVER_EXCEEDED,
                                "Number of retries on next server exceeded max " + maxRetrysNext
                                + " retries, while making a call for: " + context.getServer(), e);
                    }
                    else if (maxRetrysSame > 0 && context.getAttemptCount() == (maxRetrysSame + 1)) {
                        e = new ClientException(ClientException.ErrorType.NUMBEROF_RETRIES_EXEEDED,
                                "Number of retries exceeded max " + maxRetrysSame
                                + " retries, while making a call for: " + context.getServer(), e);
                    }
                }
                if (listenerInvoker != null) {
                    listenerInvoker.onExecutionFailed(e, context.toFinalExecutionInfo());
                }
                return Observable.error(e);
            }
        });
    }

从一组ServerList 列表中挑选合适的Server

    /**
     * Compute the final URI from a partial URI in the request. The following steps are performed:
     * <ul>
     * <li>  如果host尚未指定,则从负载均衡器中选定 host/port
     * <li>  如果host 尚未指定并且尚未找到负载均衡器,则尝试从 虚拟地址中确定host/port
     * <li> 如果指定了HOST,并且URI的授权部分通过虚拟地址设置,并且存在负载均衡器,则通过负载就均衡器中确定host/port(指定的HOST将会被忽略)
     * <li> 如果host已指定,但是尚未指定负载均衡器和虚拟地址配置,则使用真实地址作为host
     * <li> if host is missing but none of the above applies, throws ClientException
     * </ul>
     *
     * @param original Original URI passed from caller
     */
    public Server getServerFromLoadBalancer(@Nullable URI original, @Nullable Object loadBalancerKey) throws ClientException {
        String host = null;
        int port = -1;
        if (original != null) {
            host = original.getHost();
        }
        if (original != null) {
            Pair<String, Integer> schemeAndPort = deriveSchemeAndPortFromPartialUri(original);        
            port = schemeAndPort.second();
        }

        // Various Supported Cases
        // The loadbalancer to use and the instances it has is based on how it was registered
        // In each of these cases, the client might come in using Full Url or Partial URL
        ILoadBalancer lb = getLoadBalancer();
        if (host == null) {
            // 提供部分URI,缺少HOST情况下
            // well we have to just get the right instances from lb - or we fall back
            if (lb != null){
                Server svc = lb.chooseServer(loadBalancerKey);// 使用负载均衡器选择Server
                if (svc == null){
                    throw new ClientException(ClientException.ErrorType.GENERAL,
                            "Load balancer does not have available server for client: "
                                    + clientName);
                }
                //通过负载均衡器选择的结果中选择host
                host = svc.getHost();
                if (host == null){
                    throw new ClientException(ClientException.ErrorType.GENERAL,
                            "Invalid Server for :" + svc);
                }
                logger.debug("{} using LB returned Server: {} for request {}", new Object[]{clientName, svc, original});
                return svc;
            } else {
                // No Full URL - and we dont have a LoadBalancer registered to
                // obtain a server
                // if we have a vipAddress that came with the registration, we
                // can use that else we
                // bail out
                // 通过虚拟地址配置解析出host配置返回
                if (vipAddresses != null && vipAddresses.contains(",")) {
                    throw new ClientException(
                            ClientException.ErrorType.GENERAL,
                            "Method is invoked for client " + clientName + " with partial URI of ("
                            + original
                            + ") with no load balancer configured."
                            + " Also, there are multiple vipAddresses and hence no vip address can be chosen"
                            + " to complete this partial uri");
                } else if (vipAddresses != null) {
                    try {
                        Pair<String,Integer> hostAndPort = deriveHostAndPortFromVipAddress(vipAddresses);
                        host = hostAndPort.first();
                        port = hostAndPort.second();
                    } catch (URISyntaxException e) {
                        throw new ClientException(
                                ClientException.ErrorType.GENERAL,
                                "Method is invoked for client " + clientName + " with partial URI of ("
                                + original
                                + ") with no load balancer configured. "
                                + " Also, the configured/registered vipAddress is unparseable (to determine host and port)");
                    }
                } else {
                    throw new ClientException(
                            ClientException.ErrorType.GENERAL,
                            this.clientName
                            + " has no LoadBalancer registered and passed in a partial URL request (with no host:port)."
                            + " Also has no vipAddress registered");
                }
            }
        } else {
            // Full URL Case URL中指定了全地址,可能是虚拟地址或者是hostAndPort
            // This could either be a vipAddress or a hostAndPort or a real DNS
            // if vipAddress or hostAndPort, we just have to consult the loadbalancer
            // but if it does not return a server, we should just proceed anyways
            // and assume its a DNS
            // For restClients registered using a vipAddress AND executing a request
            // by passing in the full URL (including host and port), we should only
            // consult lb IFF the URL passed is registered as vipAddress in Discovery
            boolean shouldInterpretAsVip = false;

            if (lb != null) {
                shouldInterpretAsVip = isVipRecognized(original.getAuthority());
            }
            if (shouldInterpretAsVip) {
                Server svc = lb.chooseServer(loadBalancerKey);
                if (svc != null){
                    host = svc.getHost();
                    if (host == null){
                        throw new ClientException(ClientException.ErrorType.GENERAL,
                                "Invalid Server for :" + svc);
                    }
                    logger.debug("using LB returned Server: {} for request: {}", svc, original);
                    return svc;
                } else {
                    // just fall back as real DNS
                    logger.debug("{}:{} assumed to be a valid VIP address or exists in the DNS", host, port);
                }
            } else {
                // consult LB to obtain vipAddress backed instance given full URL
                //Full URL execute request - where url!=vipAddress
                logger.debug("Using full URL passed in by caller (not using load balancer): {}", original);
            }
        }
        // end of creating final URL
        if (host == null){
            throw new ClientException(ClientException.ErrorType.GENERAL,"Request contains no HOST to talk to");
        }
        // just verify that at this point we have a full URL

        return new Server(host, port);
    }


  • 发表于 2019-05-09 11:09
  • 阅读 ( 248 )
  • 分类:Java基础

条评论

请先 登录 后评论
不写代码的码农
IT 小兵

355 篇文章

作家榜 »

  1. 小编 文章
返回顶部
部分文章转自于网络,若有侵权请联系我们删除