Some of my django REST services have to connect to an asyncio server to get some information. So I'm working in a threaded environment.
While connecting, the open_connection() takes an unreasonable 2 seconds (almost exactly, always just a bit more).
Client code:
import asyncio
import datetime
def call():
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
#asyncio.coroutine
def msg_to_mars():
print("connecting", datetime.datetime.now())
reader, writer = yield from asyncio.open_connection('localhost', 8888, loop=loop)
print("connected", datetime.datetime.now()) # time reported here will be +2 seconds
return None
res = loop.run_until_complete(msg_to_mars())
loop.close()
return res
call()
Server code:
import asyncio
#asyncio.coroutine
def handle_connection(reader: asyncio.StreamReader, writer: asyncio.StreamWriter):
pass
loop = asyncio.get_event_loop()
asyncio.set_event_loop(loop)
# Each client connection will create a new protocol instance
coro = asyncio.start_server(handle_connection, '0.0.0.0', 8888, loop=loop)
server = loop.run_until_complete(coro)
# Serve requests until Ctrl+C is pressed
print('MARS Device server serving on {}'.format(server.sockets[0].getsockname()))
try:
loop.run_forever()
except KeyboardInterrupt:
pass
server.close()
loop.run_until_complete(server.wait_closed())
loop.close()
Both are basically copied from asyncio documentation samples for streamed communication, except for the additional assigning of event loop for threading.
How can I make this delay go away?
Turns out, the problem was in Windows DNS resolution.
Changing URLs from my computer name to 127.0.0.1 immediately killed the delays.
Related
Into:
I am working in a TCP server that receives events over TCP. For this task, I decided to use asyncio Protocol libraries (yeah, maybe I should have used Streams), the reception of events works fine.
Problem:
I need to be able to connect to the clients, so I create another "server" used to look up all my connected clients, and after finding the correct one, I use the Protocol class transport object to send a message and try to grab the response by reading a buffer variable that always has the last received message.
My problem is, after sending the message, I don't know how to wait for the response, so I always get the previous message from the buffer.
I will try to simplify the code to illustrate (please, keep in mind that this is an example, not my real code):
import asyncio
import time
CONN = set()
class ServerProtocol(asyncio.Protocol):
def connection_made(self, transport):
self.transport = transport
CONN.add(self)
def data_received(self, data):
self.buffer = data
# DO OTHER STUFF
print(data)
def connection_lost(self, exc=None):
CONN.remove(self)
class ConsoleProtocol(asyncio.Protocol):
def connection_made(self, transport):
self.transport = transport
# Get first value just to ilustrate
self.client = next(iter(CONN))
def data_received(self, data):
# Forward the message to the client
self.client.transport.write(data)
# wait a fraction of a second
time.sleep(0.2)
# foward the response of the client
self.transport.write(self.client.buffer)
def main():
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
loop.run_until_complete(
loop.create_server(protocol_factory=ServerProtocol,
host='0.0.0.0',
port=6789))
loop.run_until_complete(
loop.create_server(protocol_factory=ConsoleProtocol,
host='0.0.0.0',
port=9876))
try:
loop.run_forever()
except Exception as e:
print(e)
finally:
loop.close()
if __name__ == '__main__':
main()
This is not only my first experience writing a TCP server, but is also my first experience working with parallelism. So it took me days to realize that my sleep not only would not work, but I was locking the server while it "sleeps".
Any help is welcome.
time.sleep(0.2) is blocking, should not used in async programming, which will block the whole execution, if your program runing with 100 clients, the last client will be delayed for 0.2*99 seconds, which is not what you want.
the right way is trying to let program wait 0.2s but not blocking, then other concurrent clients would not be delayed,we can use thread.
import asyncio
import time
import threading
CONN = set()
class ServerProtocol(asyncio.Protocol):
def dealy_thread(self):
time.sleep(0.2)
def connection_made(self, transport):
self.transport = transport
CONN.add(self)
def data_received(self, data):
self.buffer = data
# DO OTHER STUFF
print(data)
def connection_lost(self, exc=None):
CONN.remove(self)
class ConsoleProtocol(asyncio.Protocol):
def connection_made(self, transport):
self.transport = transport
# Get first value just to ilustrate
self.client = next(iter(CONN))
def data_received(self, data):
# Forward the message to the client
self.client.transport.write(data)
# wait a fraction of a second
thread = threading.Thread(target=self.delay_thread, args=())
thread.daemon = True
thread.start()
# foward the response of the client
self.transport.write(self.client.buffer)
I've created a RPyC server. Connecting works, all my exposed methods work. Now I am looking to shut down the server from the client. Is this even possible? Security is not a concern as I am not worried about a rogue connection shutting down the server.
It is started with (Which is blocking):
from rpyc import ThreadPoolServer
from service import MyService
t = ThreadPoolServer(MyService(), port=56565)
t.start()
Now I just need to shut it down. I haven't found any documentation on stopping the server.
You can add to your Service class the method:
def exposed_stop(self):
pid = os.getpid()
if platform.system() == 'Windows':
PROCESS_TERMINATE = 1
handle = ctypes.windll.kernel32.OpenProcess(PROCESS_TERMINATE, False, pid)
ctypes.windll.kernel32.TerminateProcess(handle, -1)
ctypes.windll.kernel32.CloseHandle(handle)
else:
os.kill(pid, signal.SIGTERM)
This will make the service get its own PID and send SIGTERM to itself. There may be an better way of doing this hiding in some dark corner of the API, but I've found no better method.
If you want to do clean-up before your thread terminates, you can set up exit traps:
t = rpyc.utils.server.ThreadedServer(service, port=port, auto_register=True)
# Set up exit traps for graceful exit.
signal.signal(signal.SIGINT, lambda signum, frame: t.close())
signal.signal(signal.SIGTERM, lambda signum, frame: t.close())
t.start() # blocks thread
# SIGTERM or SIGINT was received and t.close() was called
print('Closing service.')
t = None
shutil.rmtree(tempdir)
# etc.
In case anybody is interested, I have found another way of doing it.
I'm just making the server object on a global scope, and then adding an exposed method to close it.
import rpyc
from rpyc.utils.server import ThreadedServer
class MyService(rpyc.Service):
def exposed_stop(self):
server.close()
def exposed_echo(self, text):
print(text)
server = ThreadedServer(MyService, port = 18812)
if __name__ == "__main__":
print("server start")
server.start()
print("Server closed")
On the client side, you will have an EOF error due to the connection being remotely closed. So it's better to catch it.
import rpyc
c = rpyc.connect("localhost", 18812)
c.root.echo("hello")
try :
c.root.stop()
except EOFError as e:
print("Server was closed")
EDIT: I needed to be able to dinamically specify the server. So I came with this (Is it better ? I don't know, but it works well. Be careful though, if you have multiple server running this service: things could become weird):
import rpyc
from rpyc.utils.server import ThreadedServer
class MyService(rpyc.Service):
_server:ThreadedServer
#staticmethod
def set_server(inst=ThreadedServer):
MyService._server = inst
def exposed_stop(self):
if self._server:
self._server.close()
def exposed_echo(self, text):
print(text)
if __name__ == "__main__":
server = ThreadedServer(MyService, port = 18812)
MyService.set_server(server)
print("server start")
server.start()
print("Server closed")
PS: It probably is possible to avoid the EOF error by using Asynchronous Operations
Iam trying to improve the performance of my application. It is a Python3.6 asyncio.Protocol based TCP server (SSL wrapped) handling a lot of requests.
It works fine and the performance is acceptable when only one connection is active, but as soon as another connection is opened, the client part of the application slows down. This is really noticeable once there are 10-15 client connection.
Is there a way to properly handle requests in parallel or should I resort to running multiple server instances?
/edit Added code
main.py
if __name__ == '__main__':
import package.server
server = package.server.TCPServer()
server.join()
package.server
import multiprocessing, asyncio, uvloop
asyncio.set_event_loop_policy(uvloop.EventLoopPolicy())
from package.connection import Connection
class TCPServer(multiprocessing.Process):
name = 'tcpserver'
def __init__(self, discord_queue=None):
multiprocessing.Process.__init__(self)
self.daemon = True
# some setup in here
self.start()
def run(self):
loop = uvloop.new_event_loop()
self.loop = loop
# db setup, etc
server = loop.create_server(Connection, HOST, PORT, ssl=SSL_CONTEXT)
loop.run_until_complete(server)
loop.run_forever()
package.connection
import asyncio, hashlib, os
from time import sleep, time as timestamp
class Connection(asyncio.Protocol):
connections = {}
def setup(self, peer):
self.peer = peer
self.ip, self.port = self.peer[0], self.peer[1]
self.buffer = []
#property
def connection_id(self):
if not hasattr(self, '_connection_id'):
self._connection_id = hashlib.md5('{}{}{}'.format(self.ip, self.port, timestamp()).encode('utf-8')).hexdigest()
return self._connection_id
def connection_lost(self, exception):
del Connection.connections[self.connection_id]
def connection_made(self, transport):
self.transport = transport
self.setup(transport.get_extra_info('peername'))
Connection.connections[self.connection_id] = self
def data_received(self, data):
# processing, average server side execution time is around 30ms
sleep(0.030)
self.transport.write(os.urandom(64))
The application runs on Debian 9.9 and is started via systemd
To "benchmark" I use this script:
import os, socket
from multiprocessing import Pool
from time import time as timestamp
def foobar(i):
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.connect(('127.0.0.1', 60000))
while True:
ms = timestamp()*1000
s.send(os.urandom(128))
s.recv(1024*2)
print(i, timestamp()*1000-ms)
if __name__ == '__main__':
instances = 4
with Pool(instances) as p:
print(p.map(foobar, range(0, instances)))
To answer my own question here. I went with a solution that spawned multiple instances which were listening on base_port + x and I put a nginx TCP loadbalancer in front of it.
The individual TCPServer instances are still spawned as own process and communicate among themselves via a separate UDP connection and with the main process via multiprocessing.Queue.
While this does not "fix" the problem, it provides a somewhat scalable solution for my very specific problem.
I'm using a socket to listen on a port in a while loop, with a 5 second timeout set by socket.settimeout(). But I have another method, which set's the listening port, and when called with a new port, i wanna force the socket to timeout so that I can reinitialise the socket and set the appropriate port inside the while loop. Is there a way to do that?
The socket is inside a subclass of threading.Thread
PS. Since this is my second day with Python, any other suggestions regarding any part would be most welcome. Thank you
EDIT:
I almost forgot. I want to reinitialise the socket when the setoutboundport method is called.
EDIT2
Man the whole code is messed up. I reexamined everything and it's really wrong for what I wanna achieve. Just focus on the main question. Timing out the socket.
import threading
import socket
import ResponseSender
import sys
import traceback
def __init__(self, inboundport, outboundip, outboundport, ttl=60):
super(Relay, self).__init__()
self.inboundport = inboundport
self.outboundip = outboundip
self.outboundport = outboundport
self.ttl = ttl
self.serverOn = True
self.firstpacket = True
self.sibling = None
self.newoutboundport = 0
self.listener = None
# ENDOF: __init__
def stop(self):
self.serverOn = False
# ENDOF: stop
def setsiblingrelay(self, relay):
self.sibling = relay
# ENDOF: setsiblingrelay
def setoutboundport(self, port):
self.newoutboundport = port
# ENDOF: setoutboundport
def run(self):
s = None
try:
while self.serverOn:
if not s:
s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
if self.outboundport != self.newoutboundport:
try:
s.close()
except:
pass
s.settimeout(4)
s.bind(('', self.inboundport))
print("Relay started :{0} => {1}:{2}".format(self.inboundport, self.outboundip, self.outboundport))
print("---------------------------------------- LISTENING FOR INCOMING PACKETS")
data, address = s.recvfrom(32768)
print("Received {0} from {1}:{2} => sending to {3}:{4}"
.format(data, address[0], address[1], self.outboundip, self.outboundport))
ResponseSender.sendresponse(address[0], address[1], data)
except TimeoutError:
pass
except:
print("Error: {0}".format(traceback.format_exception(*sys.exc_info())))
# ENDOF: run
I'm currently doing my first steps with asyncio in Python 3.5 and there is one problem that's bugging me. Obviously I haven't fully understood coroutines...
Here is a simplified version of what I'm doing.
In my class I have an open() method that creates a new thread. Within that thread I create a new event loop and a socket connection to some host. Then I let the loop run forever.
def open(self):
# create thread
self.thread = threading.Thread(target=self._thread)
self.thread.start()
# wait for connection
while self.protocol is None:
time.sleep(0.1)
def _thread(self):
# create loop, connection and run forever
self.loop = asyncio.new_event_loop()
coro = self.loop.create_connection(lambda: MyProtocol(self.loop),
'somehost.com', 1234)
self.loop.run_until_complete(coro)
self.loop.run_forever()
Stopping the connection is now quite simple, I just stop the loop from the main thread:
loop.call_soon_threadsafe(loop.stop)
Unfortunately I need to do some cleanup, especially I need to empty a queue before disconnecting from the server. So I tried something like this stop() method in MyProtocol:
class MyProtocol(asyncio.Protocol):
def __init__(self, loop):
self._loop = loop
self._queue = []
async def stop(self):
# wait for all queues to empty
while self._queue:
await asyncio.sleep(0.1)
# disconnect
self.close()
self._loop.stop()
The queue gets emptied from within the protocol's data_received() method, so I just want to wait for that to happen using the while loop with the asyncio.sleep() call. Afterwards I close the connection and stop the loop.
But how do I call this method from the main thread and wait for it?
I tried the following, but none of them seem to work (protocol is the currently used instance of MyProtocol):
loop.call_soon_threadsafe(protocol.stop)
loop.call_soon_threadsafe(functools.partial(asyncio.ensure_future, protocol.stop(), loop=loop))
asyncio.ensure_future(protocol.stop(), loop=loop)
Can anyone please help me here? Thanks!
Basically you want to schedule coroutine on loop of different thread. You could use run_coroutine_threadsafe:
future = asyncio.run_coroutine_threadsafe(protocol.stop, loop=loop)
future.result() # wait for results
Or the old style async like in https://stackoverflow.com/a/32084907/681044
import asyncio
from threading import Thread
loop = asyncio.new_event_loop()
def f(loop):
asyncio.set_event_loop(loop)
loop.run_forever()
t = Thread(target=f, args=(loop,))
t.start()
#asyncio.coroutine
def g():
yield from asyncio.sleep(1)
print('Hello, world!')
loop.call_soon_threadsafe(asyncio.async, g())