Python Tornado tcpserver 流写入结果

Question

我有一个简单的 tornado tcp 服务器，有时会写入客户端流。

我想得到stream.write()的结果。无论流是否仍然处于活动状态，它总是返回 None。

我已经尝试使用带有回调和协程的 stream.write()。

有没有办法实际获取 stream.write() 的结果？

这里是代码。

import logging
import signal
import tornado.httpserver
import datetime
from tornado import gen


logger = logging.getLogger(__name__)


class Server(tornado.httpserver.TCPServer):
    def __init__(self, io_loop=None, ssl_options=None, **kwargs):

        logger.debug('tcp server started')

        if not io_loop:
            io_loop = tornado.ioloop.IOLoop.instance()

        self.io_loop = io_loop
        self.streams = []

        self.io_loop.add_timeout(datetime.timedelta(seconds=5), self.write_to_all_stream)

        tornado.httpserver.TCPServer.__init__(
            self, io_loop=io_loop, ssl_options=ssl_options, **kwargs
        )        


    def handle_stream(self, stream, address):
        logger.debug('New connection from %s' % str(address))
        self.streams.append(stream)


    def on_write_stream(self):
        logger.debug('written to stream')


    #@gen.coroutine
    def write_to_all_stream(self):
        logger.debug('writing to all streams')

        for s in self.streams:
            if s.closed():
                logger.debug('Stream is closed!')
            else:
                result = s.write('hello', self.on_write_stream)
                #result = yield s.write('hello')

                logger.debug('result: %s' % str(result))


def configure_signals():
    def bye_handler(signal, frame):
        logger.info('interrupt signal received, shutting down')

        io_loop = tornado.ioloop.IOLoop.instance()
        io_loop.stop()

    signal.signal(signal.SIGINT, bye_handler)
    signal.signal(signal.SIGTERM, bye_handler)


def configure_logging():
    logging.basicConfig(
        filename=None,
        level=logging.DEBUG,
        format='%(asctime)s: %(levelname)7s: [%(name)s]: %(message)s',
        datefmt='%Y-%m-%d %H:%M:%S',
    )


if __name__ == '__main__':
    configure_signals()
    configure_logging()

    loop = tornado.ioloop.IOLoop.instance()

    logger.debug('hello')

    s = Server().listen(12345, '127.0.0.1')

    loop.start()

    logger.debug('bye')

为了测试客户端流，我使用 nc ( netcat )。我用CTRL + C/CTRL + Z模拟断开连接

nc 127.0.0.1 12345

操作系统：Ubuntu 14.04

Python：2.7.6

龙卷风版本：4.2.0

Answer 1

经过一番挖掘和测试，我找到了答案。

BaseIOStream.get_fd_error()

其中规定：

返回有关基础文件的任何错误的信息。

在 IOLoop 发出文件描述符上的错误信号后调用此方法，并应返回异常（例如带有附加信息的 socket.error，如果没有此类信息可用，则返回 None。

http://tornado.readthedocs.org/en/latest/iostream.html#tornado.iostream.BaseIOStream.get_fd_error

这是上述代码的更新版本。如果发生流错误，它会检测它们。

import logging
import signal
import tornado.httpserver
import datetime
import functools


logger = logging.getLogger(__name__)


class Server(tornado.httpserver.TCPServer):
    def __init__(self, io_loop=None, ssl_options=None, **kwargs):
        
        logger.debug('tcp server started')

        if not io_loop:
            io_loop = tornado.ioloop.IOLoop.instance()

        self.io_loop = io_loop
        self.streams = []
        
        self.io_loop.add_timeout(datetime.timedelta(seconds=5), self.write_to_all_stream)
        
        tornado.httpserver.TCPServer.__init__(
            self, io_loop=io_loop, ssl_options=ssl_options, **kwargs
        )        


    def handle_stream(self, stream, address):
        logger.debug('New connection from %s' % str(address))
        self.streams.append(stream)


    def on_write_stream(self, stream):
        status = stream.get_fd_error()
        
        if status[0]:
            logger.debug('Encountered an error when writing to stream: %s' % str(status[1]))
        else:
            logger.debug('Successfully written to stream')


    def write_to_all_stream(self):
        logger.debug('writing to all streams')
        
        for s in self.streams:
            s.write('hello', functools.partial(self.on_write_stream, s))

 
def configure_signals():
    def bye_handler(signal, frame):
        logger.info('interrupt signal received, shutting down')

        io_loop = tornado.ioloop.IOLoop.instance()
        io_loop.stop()

    signal.signal(signal.SIGINT, bye_handler)
    signal.signal(signal.SIGTERM, bye_handler)


def configure_logging():
    logging.basicConfig(
        filename=None,
        level=logging.DEBUG,
        format='%(asctime)s: %(levelname)7s: [%(name)s]: %(message)s',
        datefmt='%Y-%m-%d %H:%M:%S',
    )


if __name__ == '__main__':
    configure_signals()
    configure_logging()
    
    loop = tornado.ioloop.IOLoop.instance()
    
    logger.debug('hello')
    
    s = Server().listen(12345, '127.0.0.1')
    
    loop.start()
    
    logger.debug('bye')

更新：

stream.write() 对于 TCP 连接不是 100% 可靠的。 stream.get_fd_error() 当数据无错误地放入操作系统队列时返回 (0, 'Success') 但不保证它到达客户端。为此，您需要在应用级别实施确认系统。

更多细节在这里： https://stackoverflow.com/a/3579609/5069752