skychips
commented
7 years ago 经过了长达几个小时的抓包。 使用 Wireshark 抓包 当客户端连续5次读取数据失败（tcp retransmission） 就会与服务端主动断开连接。 这样服务端与客户端同时报出 10053 异常
I1010 17：14：01.399317 8216 tcp_conn.cc:206] evpp :: TCPConn :: HandleRead this = 0066FD30 errno = 10053您的主机中的软件中止了一个已建立的连接
void TCPConn::HandleRead() { assert(loop_->IsInLoopThread()); int serrno = 0; ssize_t n = inputbuffer.ReadFromFD(chan_->fd(), &serrno); if (n > 0) { msgfn(shared_from_this(), &inputbuffer); } else if (n == 0) { if (type() == kOutgoing) { // This is an outgoing connection, we own it and it's done. so close it DLOGTRACE << "fd=" << fd << ". We read 0 bytes and close the socket."; status_ = kDisconnecting; HandleClose(); } else { // Fix the half-closing problem : https://github.com/chenshuo/muduo/pull/117
chan_->DisableReadEvent();
if (close_delay_.IsZero()) {
DLOG_TRACE << "channel (fd=" << chan_->fd() << ") DisableReadEvent. delay time " << close_delay_.Seconds() << "s. We close this connection immediately";
DelayClose();
} else {
// This is an incoming connection, we need to preserve the
// connection for a while so that we can reply to it.
// And we set a timer to close the connection eventually.
DLOG_TRACE << "channel (fd=" << chan_->fd() << ") DisableReadEvent. And set a timer to delay close this TCPConn, delay time " << close_delay_.Seconds() << "s";
delay_close_timer_ = loop_->RunAfter(close_delay_, std::bind(&TCPConn::DelayClose, shared_from_this())); // TODO leave it to user layer close.
}
}
} else {
if (EVUTIL_ERR_RW_RETRIABLE(serrno)) {
DLOG_TRACE << "errno=" << serrno << " " << strerror(serrno);
} else {
DLOG_TRACE << "errno=" << serrno << " " << strerror(serrno) << " We are closing this connection now.";
HandleError();
}
}}
百度查阅了点资料
相关资料：TCP协议是一个可靠的协议。它通过重新发送(retransmission)来实现TCP片段传输的可靠性。简单的说，TCP会不断重复发送TCP片段，直到片段被正确接收。
还望作者指条客户端的明路。
zieckey
shgxwxl
feihongmeilian
sunwangme
I1010 17:14:01.049772 7676 eventloop.cc:289] evpp::EventLoop::QueueInLoop this=00646770 call watcher->Nofity() notified_.store(true) I1010 17:14:01.049772 8216 event_watcher.cc:154] PipeEventWatcher::HandlerFn fd=244 v=00645718 I1010 17:14:01.050272 8216 event_loop.cc:303] evpp::EventLoop::DoPendingFunctors this=00646770 pending_functorcount=1 PendingQueueSize=1 notified_=1 I1010 17:14:01.050272 8216 event_loop.cc:326] evpp::EventLoop::DoPendingFunctors this=00646770 pending_functorcount=1 PendingQueueSize=0 notified_=0 I1010 17:14:01.050272 8216 event_loop.cc:328] evpp::EventLoop::DoPendingFunctors this=00646770 pending_functorcount=1 PendingQueueSize=0 notified_=0 I1010 17:14:01.050272 8216 event_loop.cc:333] evpp::EventLoop::DoPendingFunctors this=00646770 pending_functorcount=0 PendingQueueSize=0 notified_=0 I1010 17:14:01.398816 8216 fd_channel.cc:157] evpp::FdChannel::HandleEvent this=006439C8 fd=348 kReadable
We are closing this connection now. I1010 17:14:01.399317 8216 tcp_conn.cc:287] evpp::TCPConn::HandleError this=0066FD30 fd=348 status=kConnected I1010 17:14:01.399317 8216 tcpconn.cc:247] evpp::TCPConn::HandleClose this=0066FD30 addr=(192.168.1.2:4931(local)->43.241.48.144:12345) fd=348 status=kDisconnecting I1010 17:14:01.399317 8216 fd_channel.cc:116] evpp::FdChannel::DetachFromLoop this=006439C8 fd=348 detach from event loop I1010 17:14:01.399317 8216 fd_channel.cc:28] evpp::FdChannel::Close this=006439C8 fd=348 W1010 17:14:01.399317 8216 Client.cpp:67] Disconnected from 43.241.48.144:12345 I1010 17:14:01.399317 8216 tcpconn.cc:282] evpp::TCPConn::HandleClose this=0066FD30 addr=(192.168.1.2:4931(local)->43.241.48.144:12345) fd=348 status=kDisconnecting use_count=1 I1010 17:14:01.399317 8216 tcp_conn.cc:36] evpp::TCPConn::~TCPConn this=0066FD30 name=Client channel=006439C8 fd=348 type=1 status=kDisconnected addr=(192.168.1.2:4931(local)->43.241.48.144:12345)
请问，为什么会出现这种情况。
1个小时多点，就会出现。服务端承载200+用户数量。