手机版
你好,游客 登录 注册
背景:
阅读新闻

Dubbo性能参数调优及原理

[日期:2018-05-20] 来源:Linux社区  作者:cyfonly [字体: ]

本文是针对 Dubbo 协议调用的调优指导,详细说明常用调优参数的作用域及源码。

Dubbo调用模型

常用性能调优参数

参数名 作用范围 默认值 说明 备注
threads provider 200 业务处理线程池大小  
iothreads provider CPU+1 io线程池大小  
queues provider 0

线程池队列大小,当线程池满时,排队等待执行的队列大小,

建议不要设置,当线程程池时应立即失败,

重试其它服务提供机器,而不是排队,除非有特殊需求

 
connections consumer 0

对每个提供者的最大连接数,

rmi、http、hessian等短连接协议表示限制连接数,

Dubbo等长连接协表示建立的长连接个数

Dubbo协议默认共享一个长连接
actives consumer 0 每服务消费者每服务每方法最大并发调用数 0表示不限制
acceptes provider 0 服务提供方最大可接受连接数 0表示不限制
executes provider 0 服务提供者每服务每方法最大可并行执行请求数 0表示不限制

源码及原理分析

>>  threads

FixedThreadPool.Java

public Executor getExecutor(URL url) {
    String name = url.getParameter(Constants.THREAD_NAME_KEY, Constants.DEFAULT_THREAD_NAME);
    int threads = url.getParameter(Constants.THREADS_KEY, Constants.DEFAULT_THREADS);
    int queues = url.getParameter(Constants.QUEUES_KEY, Constants.DEFAULT_QUEUES);
    return new ThreadPoolExecutor(threads, threads, 0, TimeUnit.MILLISECONDS,
            queues == 0 ? new SynchronousQueue<Runnable>() :
                    (queues < 0 ? new LinkedBlockingQueue<Runnable>() :
                            new LinkedBlockingQueue<Runnable>(queues)),
            new NamedThreadFactory(name, true), new AbortPolicyWithReport(name, url));
}

LimitedThreadPool.java

public Executor getExecutor(URL url) {
    String name = url.getParameter(Constants.THREAD_NAME_KEY, Constants.DEFAULT_THREAD_NAME);
    int cores = url.getParameter(Constants.CORE_THREADS_KEY, Constants.DEFAULT_CORE_THREADS);
    int threads = url.getParameter(Constants.THREADS_KEY, Constants.DEFAULT_THREADS);
    int queues = url.getParameter(Constants.QUEUES_KEY, Constants.DEFAULT_QUEUES);
    return new ThreadPoolExecutor(cores, threads, Long.MAX_VALUE, TimeUnit.MILLISECONDS,
            queues == 0 ? new SynchronousQueue<Runnable>() :
                    (queues < 0 ? new LinkedBlockingQueue<Runnable>() :
                            new LinkedBlockingQueue<Runnable>(queues)),
            new NamedThreadFactory(name, true), new AbortPolicyWithReport(name, url));
}

其中,Constants.DEFAULT_QUEUES = 200。threads 参数配置的是业务处理线程池的最大(或核心)线程数。

>>  iothreads

NettyServer.java

@Override
protected void doOpen() throws Throwable {
    NettyHelper.setNettyLoggerFactory();
    ExecutorService boss = Executors.newCachedThreadPool(new NamedThreadFactory("NettyServerBoss", true));
    ExecutorService worker = Executors.newCachedThreadPool(new NamedThreadFactory("NettyServerWorker", true));
    ChannelFactory channelFactory = new NioServerSocketChannelFactory(boss, worker, getUrl().getPositiveParameter(Constants.IO_THREADS_KEY, Constants.DEFAULT_IO_THREADS));
    bootstrap = new ServerBootstrap(channelFactory);
       
    final NettyHandler nettyHandler = new NettyHandler(getUrl(), this);
    channels = nettyHandler.getChannels();
    bootstrap.setPipelineFactory(new ChannelPipelineFactory() {
        public ChannelPipeline getPipeline() {
            NettyCodecAdapter adapter = new NettyCodecAdapter(getCodec() ,getUrl(), NettyServer.this);
            ChannelPipeline pipeline = Channels.pipeline();
            pipeline.addLast("decoder", adapter.getDecoder());
            pipeline.addLast("encoder", adapter.getEncoder());
            pipeline.addLast("handler", nettyHandler);
            return pipeline;
        }
    });
    // bind
    channel = bootstrap.bind(getBindAddress());
}

>>  queues

分别在 FixedThreadPool.java、LimitedThreadPool.java 和 CachedThreadPool.java 中使用,代码详情见 3.2章节。 由代码可见,默认值为 0,表示使用同步阻塞队列;如果 queues 设置为小于 0 的值,则使用容量为 Integer.MAX_VALUE 的阻塞链表队列;如果为其他值,则使用指定大小的阻塞链表队列。

>>  connections

DubboProtocol.java

private ExchangeClient[] getClients(URL url){
    //是否共享连接
    boolean service_share_connect = false;
    int connections = url.getParameter(Constants.CONNECTIONS_KEY, 0);
    //如果connections不配置,则共享连接,否则每服务每连接
    if (connections == 0){
        service_share_connect = true;
        connections = 1;
    }

    ExchangeClient[] clients = new ExchangeClient[connections];
    for (int i = 0; i < clients.length; i++) {
        if (service_share_connect){
            clients[i] = getSharedClient(url);
        } else {
            clients[i] = initClient(url);
        }
    }
    return clients;
}

DubboInvoker.java

@Override
protected Result doInvoke(final Invocation invocation) throws Throwable {
    RpcInvocation inv = (RpcInvocation) invocation;
    final String methodName = RpcUtils.getMethodName(invocation);
    inv.setAttachment(Constants.PATH_KEY, getUrl().getPath());
    inv.setAttachment(Constants.VERSION_KEY, version);
   
    ExchangeClient currentClient;
    if (clients.length == 1) {
        currentClient = clients[0];
    } else {
        currentClient = clients[index.getAndIncrement() % clients.length];
    }
    try {
        boolean isAsync = RpcUtils.isAsync(getUrl(), invocation);
        boolean isOneway = RpcUtils.isOneway(getUrl(), invocation);
        int timeout = getUrl().getMethodParameter(methodName, Constants.TIMEOUT_KEY,Constants.DEFAULT_TIMEOUT);
        if (isOneway) {
            boolean isSent = getUrl().getMethodParameter(methodName, Constants.SENT_KEY, false);
            currentClient.send(inv, isSent);
            RpcContext.getContext().setFuture(null);
            return new RpcResult();
        } else if (isAsync) {
            ResponseFuture future = currentClient.request(inv, timeout) ;
            RpcContext.getContext().setFuture(new FutureAdapter<Object>(future));
            return new RpcResult();
        } else {
            RpcContext.getContext().setFuture(null);
            return (Result) currentClient.request(inv, timeout).get();
        }
    } catch (TimeoutException e) {
        throw new RpcException(RpcException.TIMEOUT_EXCEPTION, "Invoke remote method timeout. method: " + invocation.getMethodName() + ", provider: " + getUrl() + ", cause: " + e.getMessage(), e);
    } catch (RemotingException e) {
        throw new RpcException(RpcException.NETWORK_EXCEPTION, "Failed to invoke remote method: " + invocation.getMethodName() + ", provider: " + getUrl() + ", cause: " + e.getMessage(), e);
    }
}

以上可见,默认值为0,表示针对每个 Provider,所有客户端共享一个长连接;否则,建立指定数量的长连接。在调用时,如果有多个长连接,则使用轮询方式获得一个长连接。

>>  actives

ActiveLimitFilter.java

public Result invoke(Invoker<?> invoker, Invocation invocation) throws RpcException {
    URL url = invoker.getUrl();
    String methodName = invocation.getMethodName();
    int max = invoker.getUrl().getMethodParameter(methodName, Constants.ACTIVES_KEY, 0);
    RpcStatus count = RpcStatus.getStatus(invoker.getUrl(), invocation.getMethodName());
    if (max > 0) {
        long timeout = invoker.getUrl().getMethodParameter(invocation.getMethodName(), Constants.TIMEOUT_KEY, 0);
        long start = System.currentTimeMillis();
        long remain = timeout;
        int active = count.getActive();
        if (active >= max) {
            synchronized (count) {
                while ((active = count.getActive()) >= max) {
                    try {
                        count.wait(remain);
                    } catch (InterruptedException e) {
                    }
                    long elapsed = System.currentTimeMillis() - start;
                    remain = timeout - elapsed;
                    if (remain <= 0) {
                        throw new RpcException("Waiting concurrent invoke timeout in client-side for service:  "
                                              + invoker.getInterface().getName() + ", method: "
                                              + invocation.getMethodName() + ", elapsed: " + elapsed
                                              + ", timeout: " + timeout + ". concurrent invokes: " + active
                                              + ". max concurrent invoke limit: " + max);
                    }
                }
            }
        }
    }
    try {
        long begin = System.currentTimeMillis();
        RpcStatus.beginCount(url, methodName);
        try {
            Result result = invoker.invoke(invocation);
            RpcStatus.endCount(url, methodName, System.currentTimeMillis() - begin, true);
            return result;
        } catch (RuntimeException t) {
            RpcStatus.endCount(url, methodName, System.currentTimeMillis() - begin, false);
            throw t;
        }
    } finally {
        if(max>0){
            synchronized (count) {
                count.notify();
            }
        }
    }
}

Consumer 调用时,统计服务和方法维度的调用情况,如果并发数超过设置的最大值,则阻塞当前线程,直到前面有请求处理完成。

>>  accepts

AbstractServer.java

@Override
public void connected(Channel ch) throws RemotingException {
    Collection<Channel> channels = getChannels();
    if (accepts > 0 && channels.size() > accepts) {
        logger.error("Close channel " + ch + ", cause: The server " + ch.getLocalAddress() + " connections greater than max config " + accepts);
        ch.close();
        return;
    }
    super.connected(ch);
}

当连接数大于最大值时,关闭当前连接。

>>  executes

ExecuteLimitFilter.jvava

public Result invokeOrg(Invoker<?> invoker, Invocation invocation) throws RpcException {
    URL url = invoker.getUrl();
    String methodName = invocation.getMethodName();
    int max = url.getMethodParameter(methodName, Constants.EXECUTES_KEY, 0);
    if (max > 0) {
        RpcStatus count = RpcStatus.getStatus(url, invocation.getMethodName());
        if (count.getActive() >= max) {
            throw new RpcException("Failed to invoke method " + invocation.getMethodName() + " in provider " + url + ", cause: The service using threads greater than <dubbo:service executes=\"" + max + "\" /> limited.");
        }
    }
    long begin = System.currentTimeMillis();
    boolean isException = false;
    RpcStatus.beginCount(url, methodName);
    try {
        Result result = invoker.invoke(invocation);
        return result;
    } catch (Throwable t) {
        isException = true;
        if(t instanceof RuntimeException) {
            throw (RuntimeException) t;
        }
        else {
            throw new RpcException("unexpected exception when ExecuteLimitFilter", t);
        }
    }
    finally {
        RpcStatus.endCount(url, methodName, System.currentTimeMillis() - begin, isException);
    }
}

Provider处理请求时,统计方法维度的调用情况,如果并发数超过设置的最大值,则阻直接抛出异常。

本文永久更新链接地址https://www.linuxidc.com/Linux/2018-05/152467.htm

linux
本文评论   查看全部评论 (0)
表情: 表情 姓名: 字数

       

评论声明
  • 尊重网上道德,遵守中华人民共和国的各项有关法律法规
  • 承担一切因您的行为而直接或间接导致的民事或刑事法律责任
  • 本站管理人员有权保留或删除其管辖留言中的任意内容
  • 本站有权在网站内转载或引用您的评论
  • 参与本评论即表明您已经阅读并接受上述条款