python SimpleHTTPRequestHandler 服务器在退出后将套接字留在 TIME_WAIT 状态

Question

我编写了一个简单的服务器来扩展 SimpleHTTPRequestHandler

如果我在不向服务器发出任何请求的情况下启动和停止它，我可以毫无问题地在同一端口上重新启动。

启动时，netstat 如下所示:

sam@hersheezy:server$ sudo netstat -na --program | grep 8001
tcp        0      0 0.0.0.0:8001            0.0.0.0:*               LISTEN      23392/python

After a request is made, netstat looks like this (even after the request has completed):

sam@hersheezy:server$ sudo netstat -na --program | grep 8001
tcp        0      0 0.0.0.0:8001            0.0.0.0:*               LISTEN      23392/python    
tcp        0      0 127.0.0.1:8001          127.0.0.1:48659         TIME_WAIT   -

Then, I kill the server using C-c and netstat looks like this (at this point I cannot restart the server because port is already in use):

 sudo netstat -na --program | grep 8001
tcp        0      0 127.0.0.1:8001          127.0.0.1:48674         TIME_WAIT   - 

I am obviously not closing something correctly. My code that sends the reply looks like the following:

"""
reply is an object that can be json encoded that is written with a response code 200
"""
def send_provider_reply(self, replyobj):
    try:
        str_reply = json.dumps(replyobj)
        self.send_response(200)
        self.send_header('Content-type', 'application/json')
        self.end_headers()
        #do we need to send a newline??
        self.wfile.write(str_reply)
    except:
        traceback.print_exc()
        self.send_err(500, 'failed after provider creation')

Answer 1

套接字选项 SO_LINGER 确实阻止套接字进入 TIME_WAIT。但是 TIME_WAIT 存在是有原因的:它应该保护您免受来自旧连接的挥之不去的数据包的影响。因此，TIME_WAIT 的默认持续时间是网络往返的两倍。所以，在 TIME_WAIT 中找到一些较旧的连接是正常的。

给出一些上下文:在服务器端，对于监听套接字，有 SO_REUSEADDR 套接字选项。它允许监听套接字在 TIME_WAIT 结束之前成功绑定(bind)。对于始终应该监听相同端口的服务器进程(想想:网络服务器在端口 80、443)，这是必须的。

服务器的典型 Python 代码可能包含如下内容:

...
listener = socket(AF_INET, SOCK_STREAM)
listener.setsockopt(SOL_SOCKET, SO_REUSEADDR, 1)
listener.bind((HOST, PORT))
listener.listen(32)
...