我已经在python中实现了一个负载平衡脚本，它将流量引导到我网络中的服务器，并在没有可用的情况下实现回退。它有效，但问题是它运行缓慢。就像非常慢。通常需要0.8秒才能加载的页面需要13秒才能加载此脚本。
#!/usr/bin/env python
# This is a simple load balancing python script, used to reduce the load on my raspberry pi servers.
import socket, select, time, sys, os, multiprocessing
def checkAdmin():
import ctypes
try:
is_admin = os.getuid() == 0
except AttributeError:
is_admin = ctypes.windll.shell32.IsUserAnAdmin() != 0
return is_admin
if not checkAdmin():
print 'ERROR: You need to run this script with admin/root permissions'
sys.exit(0)
# Changing the buffer_size and delay, you can improve the speed and bandwidth.
# But when buffer get to high or delay go too down, you can broke things
buffer_size = 4096
delay = 0.0001
forwards = {
80: {
'example.com' : {
'forwardifavailable':[
('192.168.1.3',80)
('192.168.1.4',80)
('192.168.1.5',80)
('192.168.1.6',80)
],
'forwardifunavailable':[
('127.0.0.1',1337)
],
'last_svr':0
},
'subdomain.example.com' : {
'forwardifavailable':[
('192.168.1.3',80)
('192.168.1.5',80)
('192.168.1.8',8080)
],
'forwardifunavailable':[
('127.0.0.1',1337)
],
'last_svr':0
},
'example2.com' : {
'forwardifavailable':[
('192.168.1.3',80)
],
'forwardifunavailable':[
('127.0.0.1',1337)
],
'last_svr':0
},
'unknown' : {
'forwardifavailable':[
('192.168.1.3',80)
],
'forwardifunavailable':[
('127.0.0.1', 1337)
],
'last_svr':0
}
},
8080: {
'example.com' : {
'forwardifavailable':[
('192.168.1.3',80)
('192.168.1.4',80)
('192.168.1.5',80)
('192.168.1.6',80)
],
'forwardifunavailable':[
('127.0.0.1',1337)
],
'last_svr':0
},
'subdomain.example.com' : {
'forwardifavailable':[
('192.168.1.3',80)
('192.168.1.5',80)
('192.168.1.8',8080)
],
'forwardifunavailable':[
('127.0.0.1',1337)
],
'last_svr':0
},
'example2.com' : {
'forwardifavailable':[
('192.168.1.3',80)
],
'forwardifunavailable':[
('127.0.0.1',1337)
],
'last_svr':0
},
'unknown' : {
'forwardifavailable':[
('192.168.1.3',80)
],
'forwardifunavailable':[
('127.0.0.1', 1337)
],
'last_svr':0
}
}
}
available = (multiprocessing.Manager()).dict()
def checkserver(address,port):
global available
available[(address,port)] = False
while True:
s = socket.socket()
try:
s.connect((address, port))
available[(address,port)] = True
except socket.error, e:
available[(address,port)] = False
time.sleep(5) # 5 Second wait before second check; Reduces load and makes this much faster
sockets = []
def set_up_servers():
ipandports = set([])
global sockets
for xport,sites in forwards.iteritems():
currsocket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
currsocket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR,1)
currsocket.bind(('',xport))
currsocket.listen(200)
sockets.append(currsocket)
for site,forwardslist in sites.iteritems():
for xtuple in forwardslist['forwardifavailable']:
ipandports.add(xtuple)
# Now we have a list of ips and ports we need to constantly check if up
for xtuple in ipandports:
print xtuple
(multiprocessing.Process(
target=checkserver,
args=xtuple
)).start()
def getForwardForUrl(originalurl,port, path):
starttime = time.time()
import tldextract
url = originalurl
xtuple = tuple([url,port])
subdomain = ''
if ':' in url: # In order to correctly handle ports
url = url.split(':')[0]
if url in forwards[port]: # This is to, optionally, support a subdomain on a separate server
xchksvr = forwards[port][url]
else:
extracteddomain = tldextract.extract(url)
url = "{0}.{1}".format(extracteddomain.domain,extracteddomain.suffix)
subdomain = extracteddomain.subdomain + '.'
if url in forwards[port]:
xchksvr = forwards[port][url]
else:
print 'Unknown domain: ' + originalurl + " with port: " + str(port)
xchksvr = forwards[port]['unknown']
i = len(xchksvr['forwardifavailable'])
seci = (xchksvr['last_svr'] + 1) % len(xchksvr['forwardifavailable'])
while (i > 0):
if available[xchksvr['forwardifavailable'][seci]]:
xchksvr['last_svr'] = seci
print "User requested site {0} (path: {1}) from port {2}. Directing user to server {3} with port {4}.".format(originalurl,path,port,xchksvr['forwardifavailable'][seci][0],xchksvr['forwardifavailable'][seci][1])
return xchksvr['forwardifavailable'][seci]
seci = (seci + 1) % len(xchksvr['forwardifavailable'])
i -= 1
xchksvr['last_svr'] = (xchksvr['last_svr'] + 1) % len(xchksvr['forwardifunavailable'])
print "No servers available for site {0} (path: {1}) from port {2}. Directing to fallback server {3} with port {4}.".format(originalurl,path,port,xchksvr['forwardifunavailable'][xchksvr['last_svr']][0],xchksvr['forwardifunavailable'][xchksvr['last_svr']][1])
print str(time.time() - starttime)
return xchksvr['forwardifunavailable'][xchksvr['last_svr']]
class Forward:
def __init__(self):
self.forward = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
def start(self, addresstuple):
try:
self.forward.connect(addresstuple)
return self.forward
except Exception, e:
print e
return False
class TheServer:
input_list = []
channel = {}
def __init__(self):
pass
def main_loop(self,listeningsocket):
self.input_list.append(listeningsocket)
while 1:
time.sleep(delay)
try:
ss = select.select
inputready, outputready, exceptready = ss(self.input_list, [], [])
for self.s in inputready:
if self.s == listeningsocket: # If this is an externally listening socket...
self.on_accept()
break
try:
self.data = self.s.recv(buffer_size)
except Exception:
data=''
if len(self.data) == 0:
self.on_close()
break
else:
self.on_recv(self.data)
except Exception,e:
pass
def on_accept(self):
forward = (Forward()).start(('127.0.0.1', 1337))
clientsock, clientaddr = self.s.accept()
if forward:
self.input_list.append(clientsock)
self.input_list.append(forward)
self.channel[clientsock] = forward
self.channel[forward] = clientsock
else:
print "Can't establish connection with remote server."
clientsock.close()
def on_close(self):
#remove objects from input_list
self.input_list.remove(self.s)
if self.channel[self.s] in self.input_list: # Because I'm too lazy to fix a bug where the element doesn't exist sometimes
self.input_list.remove(self.channel[self.s])
out = self.channel[self.s]
# close the connection with client
self.channel[out].close() # equivalent to do self.s.close()
# close the connection with remote server
self.channel[self.s].close()
# delete both objects from channel dict
del self.channel[out]
del self.channel[self.s]
def parse_data(self,data):
#print data # used for debugging
if (data.startswith("HTTP 1/1")):
# Data sent from server
pass
else:
# Data sent to server
if (data.startswith("GET") or (data.startswith("POST")) or (data.startswith("PUT")) or (data.startswith("PATCH")) or (data.startswith("HEAD")) or (data.startswith("OPTIONS")) or (data.startswith("DELETE"))):
# GET/POST/PUT/PATCH/HEAD/OPTIONS/DELETE request; Check that if we need a special server for this URL
timestart = time.time()
newChannel = Forward().start(getForwardForUrl(data.splitlines()[1].split()[1], self.s.getsockname()[1], data.splitlines()[0].split()[1]))
self.input_list[self.input_list.index(self.s)] = newChannel
oldForward = self.channel[self.s]
self.channel[self.s] = newChannel
del self.channel[oldForward]
self.channel[newChannel] = self.s
self.input_list.remove(oldForward)
self.input_list.append(newChannel)
self.input_list.remove(self.channel[self.s])
return data
def on_recv(self,data):
# here we can parse and/or modify the data before send forward
# data can NOT be gzipped, or this will not work
data = self.parse_data(data)
self.channel[self.s].send(data)
if __name__ == '__main__':
import signal
set_up_servers()
for xsocket in sockets:
print "Creating server to listen on port ", xsocket.getsockname()[1]
server = TheServer()
(multiprocessing.Process(target=server.main_loop,args=(xsocket,))).start()
while True:
try:
pass
except KeyboardInterrupt:
print "Ctrl C - Stopping server"
os.kill(os.getpid(), signal.SIGTERM)
我尝试进行一些代码分析，但是一旦多处理模块启动第二个进程，分析就不会对该进程进行分析，导致缺少这些代码行的分析信息。任何帮助，将不胜感激。谢谢！
BTW：这是在Raspberry Pi 2 B上运行的，与Raspbian Jessie一起。