C ++关闭Boost ASIO SSL套接字的最佳方法是什么？

Question

I have a client application that uses Boost ASIO SSL sockets. I am running into connection errors that leave the socket in an unusable state until I restart the app. To begin, I make a connection request with this code:

void apiClient::connect(boost::asio::ip::tcp::resolver::iterator endpoint_list)
{
    boost::asio::ip::tcp::resolver::iterator endpoint_iter = endpoint_list;
    boost::asio::ip::tcp::endpoint endpoint = *endpoint_iter;

    if (!ssl_socket_->lowest_layer().is_open()) {
        ssl_socket_->lowest_layer().async_connect(endpoint,
            boost::bind(&apiClient::handle_connect, this,
            boost::asio::placeholders::error, ++endpoint_iter));
    }
}

Periodically I encounter an error returned to the connection handler, the error in ec.message is [A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond] Error code [10060]

Here is my connection handler:

void apiClient::handle_connect(const boost::system::error_code& ec, 
    boost::asio::ip::tcp::resolver::iterator endpoint_list)
{
    bool found_error(false);

    if (!ec) {
        ssl_socket_->async_handshake(boost::asio::ssl::stream_base::client,
            boost::bind(&apiClient::send_request, this, 
            boost::asio::placeholders::error));
    }
    else if (endpoint_list != boost::asio::ip::tcp::resolver::iterator()) {
        std::wstring sMsg = boost::str(boost::wformat(L"API server connection failure to host [%s]. Trying next endpoint") % s2ws(endpoint_list->host_name()));
        LogMsg(sMsg);
        found_error = true;
        epList = endpoint_list;
    }
    else {
        std::wstring sMsg = boost::str(boost::wformat(L"API server connection failure.  Error msg [%s] Error code [%d]") % s2ws(ec.message()) % ec.value());
        LogMsg(sMsg);
        found_error = true;
        epList = start_endpoint_list;
    }

    if (found_error) {
        requeue_request(SHUTDOWN_CONN);
    }
}

In my requeue_request function I check the number of errors I have encountered for the request and if it exceeds a threshold, I discard the request and move on. If I have not exceeded the threshold, I shutdown the socket connection before attempting to reconnect and send the message again. My understanding is that the emplace method recreates the socket instance.

void apiClient::reconnect(SHUTDOWN_CONN_ACTION action)
{
    if (action == SHUTDOWN_CONN) {
        shutdownSocket();
    }

    ssl_socket_.emplace(*io_context_, *ctx_);

    connect_timer.expires_from_now(boost::posix_time::milliseconds(3000));
    connect_timer.async_wait(boost::bind(&apiClient::handleConnectTimer, this, boost::asio::placeholders::error));
}

void CNXGHotKeyAPIClient::shutdownSocket()
{
    boost::system::error_code ec;

    ssl_socket_->next_layer().cancel(ec);
    ssl_socket_->shutdown(ec);
    ssl_socket_->lowest_layer().shutdown(boost::asio::ip::tcp::socket::shutdown_both, ec);
    ssl_socket_->lowest_layer().close(ec);
}

This design is not working for me. When I encounter any connection error, my application retries the connection 3 times, it fails with the same error all three times and the socket is left in a state where I cannot use it until I restart my application. The server I'm connecting to is honoring hundreds of connection requests from python apps that do not use Boost so I don't think the server is at fault.

I have tried to find examples that use async_shutdown but I cannot find anything that helps me. The example at this link What is the proper way to securely disconnect an asio SSL socket?

uses the following code which I am unable to make work. example code:

boost::system::error_code ec;
ssl_socket.cancel(ec);
ssl_socket.async_shutdown([](...) { ssl_socket.close(); };

ssl_socket.cancel() doesn't even exist in my version of Boost (v1.66) the closest thing I can find is ssl_socket.next_layer().cancel(ec)

My question is; how do you shutdown a socket after it has encountered an error so that it is not left in an unusable state? When my application encounters connection errors, many times I cannot reconnect until I restart the app. The app is still running and servicing user actions so I know it has not crashed, but the socket is in a state that is unusable. Thanks for any help

Answer 1

1. You can't shutdown an SSL socket. You have to close it.
2. You can't reconnect a socket after it has been connected, even if the connect attempt failed. You have to close it and create a new one.