使用案例

GraphDaemon.cpp

/*
DEFINE_int32(num_netio_threads, 0,
                "Number of networking threads, 0 for number of physical CPU cores");
DEFINE_int32(num_accept_threads, 1, "Number of threads to accept incoming connections");
DEFINE_int32(num_worker_threads, 0, "Number of threads to execute user queries");
DEFINE_bool(reuse_port, true, "Whether to turn on the SO_REUSEPORT option");
DEFINE_int32(listen_backlog, 1024, "Backlog of the listen socket");
*/

   auto threadFactory = std::make_shared<folly::NamedThreadFactory>("graph-netio");
    auto ioThreadPool = std::make_shared<folly::IOThreadPoolExecutor>(
                            FLAGS_num_netio_threads, std::move(threadFactory));
    gServer = std::make_unique<apache::thrift::ThriftServer>();
    gServer->setIOThreadPool(ioThreadPool);

    auto interface = std::make_shared<GraphService>();
    status = interface->init(ioThreadPool);
    if (!status.ok()) {
        LOG(ERROR) << status;
        return EXIT_FAILURE;
    }

    gServer->setInterface(std::move(interface));
    gServer->setAddress(localIP, FLAGS_port);
    // fbthrift-2018.08.20 always enables SO_REUSEPORT once `setReusePort' is called
    // which had been fixed in later version.
    if (FLAGS_reuse_port) {
        gServer->setReusePort(FLAGS_reuse_port);
    }
    gServer->setIdleTimeout(std::chrono::seconds(FLAGS_client_idle_timeout_secs));
    gServer->setNumCPUWorkerThreads(FLAGS_num_worker_threads);
    gServer->setCPUWorkerThreadName("executor");
    gServer->setNumAcceptThreads(FLAGS_num_accept_threads);
    gServer->setListenBacklog(FLAGS_listen_backlog);
    gServer->setThreadStackSizeMB(5);

    FLOG_INFO("Starting nebula-graphd on %s:%d\n", localIP.c_str(), FLAGS_port);
    try {
        gServer->serve();  // Blocking wait until shut down via gServer->stop()
    } catch (const std::exception &e) {
        FLOG_ERROR("Exception thrown while starting the RPC server: %s", e.what());
        return EXIT_FAILURE;
    }

Thrift中的工作模型

• apache的Thrift（二）：各种 Server 实现
• Fbthrift: 单acceptor + 多io线程池 + 多业务线程池

ThriftServer.h 相关option设置

• ThriftServer服务端，关于其部分方法说明可参考: ThriftServer

• 配置服务端的处理器: setInterface， 每次新的请求都使用同一个实例来处理。参考：AsyncProcessorFactory 的getProcessor返回一个新的实例

可选：
auto handler = std::make_shared<ServiceHandler>(kvStore);
auto proc_factory = std::make_shared<ThriftServerAsyncProcessorFactory<ServiceHandler>>(
                           handler);
server->setProcessorFactory(proc_factory);

• ip，端口 : setAddress

• setIOThreadPool: 配置IO线程池， 线程池用来处理网络IO任务，从连接中读取请求数据并反序列化，然后就交给CPUWorker处理，将结果数据序列化并通过连接发送出去。

• setNumCPUWorkerThreads： 配置cpu线程数，负责处理用户的业务逻辑，如果请求过多可以调大该值，和线程名称：用来处理非IO任务，比如计算任务：setNumCPUWorkerThreads、setCPUWorkerThreadName。 setNumCPUWorkerThreads和setThreadManager是互斥，只能选择设置一个。因为setNumCPUWorkerThreads设置后，内部会创建ThreadManager。

serve方法中：根据setNumCPUWorkerThreads的结果（getNumCPUWorkerThreads）
创建ThreadManager然后调用setThreadManager。

void ThriftServer::setupThreadManager() {
  if (!threadManager_) {
    std::shared_ptr<apache::thrift::concurrency::ThreadManager> threadManager(
        PriorityThreadManager::newPriorityThreadManager(
            getNumCPUWorkerThreads(), true /*stats*/));
    threadManager->enableCodel(getEnableCodel());
    // If a thread factory has been specified, use it.
    if (threadFactory_) {
      threadManager->threadFactory(threadFactory_);
    }
    auto poolThreadName = getCPUWorkerThreadName();
    if (!poolThreadName.empty()) {
      threadManager->setNamePrefix(poolThreadName);
    }
    threadManager->start();
    setThreadManager(threadManager);
  }
}

• setIdleTimeout：配置客户端连接成功后，如果连接上没有请求空闲到该值后，则断开连接，默认一直等待：
• setListenBacklog: 配置socket的全连接队列大小，但是需要考虑系统参数net.core.somaxconn的值（默认50），最终socket的全连接队列大小选其中小的一个，当同一个时间点同时出现的连接数很大时，可以调高这两个值：setListenBacklog

# The number of threads to execute user queries, 0 for # of CPU cores
gServer->setNumCPUWorkerThreads(FLAGS_num_worker_threads);
gServer->setCPUWorkerThreadName("executor");

# The number of threads to accept incoming connections
gServer->setNumAcceptThreads(FLAGS_num_accept_threads);