我想測試客戶端程式和伺服器程式之間的互動以了解競爭條件。他們透過 TCP 相互連接。客戶端在多個執行緒上阻塞對伺服器的呼叫(每個執行緒一個到伺服器的 TCP 連線)。我想測試一個阻塞呼叫先於另一個呼叫完成的競爭條件。
為此,我希望將 tcp 連線延遲不同的量(這樣我就不必明確重寫客戶端或伺服器)。
我希望做這樣的事情:
socat tcp-listen:$pin system:"delaycat $delay |socat - 'tcp:localhost:$pout'"
其中$pin
是客戶端連接的端口,$delay
是延遲秒數,$pout
是伺服器偵聽的端口。 Anddelaycat
是一個虛構的程序,它將輸入流延遲 n 秒。
是否有一些現有的程序可以滿足我的要求?還是我應該寫delaycat
?
編輯:請注意,系統範圍的延遲實際上並不理想,因為如果可能的話,我希望將各個套接字延遲不同的量。
答案1
我找不到任何東西,所以我寫了一個 python 腳本來做到這一點(這可能是有問題的):
#!/usr/bin/env python26
"""Add latency to a tcp connection"""
from argparse import ArgumentParser, ArgumentDefaultsHelpFormatter
from functools import partial
from twisted.internet.reactor import (
listenTCP, connectTCP, callLater, run
)
from twisted.internet.protocol import (
ServerFactory, ReconnectingClientFactory, Protocol
)
class ConnectionToServer(Protocol):
def __init__(self, connection_from_client, delay_to_client):
self._delay_to_client = delay_to_client
self._connection_from_client = connection_from_client
def connectionMade(self):
self._connection_from_client.connection_to_server = self
if self._connection_from_client.buffer_from_client:
self.transport.write(
self._connection_from_client.buffer_from_client
)
def dataReceived(self, data):
callLater(
self._delay_to_client,
self._connection_from_client.transport.write, data
)
def connectionLost(self, reason):
callLater(
self._delay_to_client,
self._connection_from_client.transport.loseConnection
)
class ConnectionFromClient(Protocol):
def __init__(self, server_host, server_port, delay_to_client, delay_to_server):
self._delay_to_server = delay_to_server
self.connection_to_server = None
self.buffer_from_client = ''
server_connection_factory = ReconnectingClientFactory()
server_connection_factory.protocol = partial(
ConnectionToServer, self, delay_to_client
)
self._server_connector = connectTCP(
server_host, server_port, server_connection_factory
)
def dataReceived(self, data):
callLater(self._delay_to_server, self._write, data)
def connectionLost(self, reason):
callLater(
self._delay_to_server, self._server_connector.disconnect
)
def _write(self, data):
if self.connection_to_server:
self.connection_to_server.transport.write(data)
else:
self.buffer_from_client += data
def main():
"""Add latency to a tcp connection"""
parser = ArgumentParser(
description=main.__doc__,
formatter_class=ArgumentDefaultsHelpFormatter
)
parser.add_argument(
'client_port', type=int, help='client connects to this port'
)
parser.add_argument(
'server_port', type=int, help='server listens on this port'
)
parser.add_argument(
'-t', '--server-host', default='localhost',
help='server is running on this host'
)
parser.add_argument(
'-c', '--delay-to-client', default=0, type=float,
help='messages to client are delayed by this many seconds'
)
parser.add_argument(
'-s', '--delay-to-server', default=0, type=float,
help='messages to server are delayed by this many seconds'
)
args = parser.parse_args()
client_connection_factory = ServerFactory()
client_connection_factory.protocol = partial(
ConnectionFromClient, args.server_host, args.server_port,
args.delay_to_client, args.delay_to_server
)
listenTCP(args.client_port, client_connection_factory)
run()
if __name__ == '__main__':
main()
答案2
在 Linux 機器上,您可以使用 tc內特姆。
例如,該命令tc qdisc add dev eth0 root netem delay 100ms
將延遲全部傳出資料包延遲 100 毫秒。要延遲傳入資料包,您可以使用ifb - 中間功能塊