Python's SimpleHTTPServer started in a thread won't close the port

Question

I have the following code:

import os
from ghost import Ghost
import urlparse, urllib

import SimpleHTTPServer
import SocketServer

import sys, traceback

from threading import Thread, Event
from time import sleep

please_die = Event() # this is my enemy

httpd = None
PORT = 8001
address = 'http://localhost:'+str(PORT)+'/'
search_dir = './category'

def main():
    """
      basic run script routine, 
      FIXME: is supossed to exits gracefully
    """
    thread = Thread(target = simpleServe)
    try:
      thread.start()
      run()
    except KeyboardInterrupt:
      print "Shutdown requested"
    except Exception:
      traceback.print_exc(file=sys.stdout)

    shutdown()
    sys.exit(0)

def shutdown():
  global httpd
  global please_die
  print "Shutting down"
  # A try - except for the shutdown routine
  try:
    please_die.wait() # how do you do? 

    httpd.shutdown() # Please! I whant to run you multiple times. 
    print "Have you died?"
  except Exception:
    traceback.print_exc(file=sys.stdout)

def path2url(path):
  """
  constructs an url from a relative path / concatenates the global address
  variable with the path given
  """
  global address
  return urlparse.urljoin(address, urllib.pathname2url(path))

def simpleServe():
  global httpd, PORT

  please_die.set() # Attaching the event to this thread

  # Start the service
  Handler = SimpleHTTPServer.SimpleHTTPRequestHandler

  httpd = SocketServer.TCPServer(("", PORT), Handler)

  print "serving at port", PORT
  # And loop infinetly in the hope that I can stop you later
  httpd.serve_forever()

def run():
  global search_dir;

  ghost = Ghost() # the webkit facade

  with ghost.start() as session:

    session.set_viewport_size(2560, 1600) # "retina" size

    for directory, subdirectories, files in os.walk(search_dir):
        for file in files:
            path = os.path.join(directory, file)
            urlPath = path2url(path)
            process(session, urlPath);

def process(session, urlPath):
  page, resources = session.open(urlPath)
  assert page.http_status == 200
  # ... other asserts here 


if __name__ == '__main__':
  main()

The idea is to make a script that starts a "simple http server", do some requests on it and then exit.

First time it runs without any problems:

...
127.0.0.1 - - [31/Jul/2015 13:16:17] "GET /category/52003.html HTTP/1.1" 200 -
127.0.0.1 - - [31/Jul/2015 13:16:17] "GET /category/52003.html HTTP/1.1" 200 -
127.0.0.1 - - [31/Jul/2015 13:16:17] "GET /category/52003.html HTTP/1.1" 200 -
127.0.0.1 - - [31/Jul/2015 13:16:17] "GET /static/img/glyphicons-halflings.png HTTP/1.1" 200 -
Shutting down
Have you died?

Launching it the second time crashes saying that the:

Address already in use

Exception in thread Thread-1:
Traceback (most recent call last):
  File "/usr/lib/python2.7/threading.py", line 810, in __bootstrap_inner
    self.run()
  File "/usr/lib/python2.7/threading.py", line 763, in run
    self.__target(*self.__args, **self.__kwargs)
  File "download-images.py", line 51, in simpleServe
    httpd = SocketServer.TCPServer(("", PORT), Handler)
  File "/usr/lib/python2.7/SocketServer.py", line 420, in __init__
    self.server_bind()
  File "/usr/lib/python2.7/SocketServer.py", line 434, in server_bind
    self.socket.bind(self.server_address)
  File "/usr/lib/python2.7/socket.py", line 228, in meth
    return getattr(self._sock,name)(*args)
error: [Errno 98] Address already in use

If I kill all python processes than the script runs again, and because of that I'm assuming that I used the thread wrong, but I cannot find where.

Update

Forgot to mention that,

my OS is :

$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description:    Ubuntu 15.04
Release:        15.04
Codename:       vivid

The python that I'm using is :

$ python --version
Python 2.7.9

$ netstat -putelan | grep 8001 prints :

$ netstat -putelan | grep 8001
(Not all processes could be identified, non-owned process info
    cp        0      0 127.0.0.1:34691         127.0.0.1:8001          TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:8001          127.0.0.1:34866         TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:34798         127.0.0.1:8001          TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:8001          127.0.0.1:34588         TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:34647         127.0.0.1:8001          TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:34915         127.0.0.1:8001          TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:34674         127.0.0.1:8001          TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:34451         127.0.0.1:8001          TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:8001          127.0.0.1:34930         TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:8001          127.0.0.1:34606         TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:34505         127.0.0.1:8001          TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:34717         127.0.0.1:8001          TIME_WAIT   0          0           -               
    tcp        0      0 127.0.0.1:8001          127.0.0.1:34670         0      0 127.0.0.1:8001          127.0.0.1:34626         
...

I can't post the whole sequence (due to the post limits of stackoverflow). The rest is the same with 34*** port mixed with 8001 port in an uniform sequence.

http://stackoverflow.com/questions/16730771/python-basehttpserver-httpserver-callback-for-start-and-stop-event — dsgdfg, Jul 31 '15 at 11:55
you can simply try to add `import signal try: os.kill(%d, signal.SIGTERM)%PORT` before launching server to kill activity on required port — Andersson, Jul 31 '15 at 12:10
Can you please post the output of the command `netstat -putelan | grep 8001` ? — FunkySayu, Jul 31 '15 at 12:46
@Andersson no, I would like a graceful shut down instead of a brute force kill at startup, the port is somehow configurable — Tiberiu C., Jul 31 '15 at 13:30
@FunkySayu, no it didn't, I think I understand what's happening and I thank you for that but the "flag" suggest by you was invalid in my code. — Tiberiu C., Jul 31 '15 at 13:50

score 15 · Accepted Answer · edited Feb 28 '16 at 12:02

As @LFJ say, this is probably due to the allow_reuse_address attribute of the TCPServer.

httpd = SocketServer.TCPServer(("", PORT), Handler, bind_and_activate=False)
httpd.allow_reuse_address = True

try:
    httpd.server_bind()
    httpd.server_activate()
except:
    httpd.server_close()
    raise

Equivalent code :

SocketServer.TCPServer.allow_reuse_address = True
https = SocketServer.TCPServer(("", PORT), Handler)

Let's explain a bit why.

When you enable TCPServer.allow_reuse_address, it adds an option on the socket :

class TCPServer:
    [...]
    def server_bind(self):
        if self.allow_reuse_address:
            self.socket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
        [...]

What is socket.SO_REUSEADDR ?

This socket option tells the kernel that even if this port is busy (in
the TIME_WAIT state), go ahead and reuse it anyway.  If it is busy,
but with another state, you will still get an address already in use
error.  It is useful if your server has been shut down, and then
restarted right away while sockets are still active on its port.  You
should be aware that if any unexpected data comes in, it may confuse
your server, but while this is possible, it is not likely.

In fact, it allows the reuse of your socket socket binding address. If another process try to bind while the socket is not listening, the process will be allowed to use this socket binding address.

The reason you need to enable that is because you don't shutdown properly your TCPServer. In order to close it properly, you must run shutdown method, which will close the thread launched by server_forever and then close the socket properly by calling the server_close method.

def shutdown():
    global httpd
    global please_die
    print "Shutting down"

    try:
        please_die.wait() # how do you do? 
        httpd.shutdown() # Stop the serve_forever
        httpd.server_close() # Close also the socket.
    except Exception:
        traceback.print_exc(file=sys.stdout)

Fujiao Liu · Answer 2 · 2015-07-31T13:48:13.917

4

i saw the TCPServer source code:

def server_bind(self):
        """Called by constructor to bind the socket.

        May be overridden.

        """
        if self.allow_reuse_address:
            self.socket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
        self.socket.bind(self.server_address)
    self.server_address = self.socket.getsockname()

allow_reuse_address should be set befor bind. so try this:

SocketServer.TCPServer.allow_reuse_address=True
httpd = SocketServer.TCPServer(("", PORT), Handler)

edited Jul 31 '15 at 13:48

answered Jul 31 '15 at 12:45

Fujiao Liu

2,195
2
24
28

Whooho, you got it! :) – Tiberiu C. Jul 31 '15 at 13:46
4

Remind me to never try to explain properly the cause and consequences of things. – FunkySayu Jul 31 '15 at 13:49

score 2 · Answer 3 · answered Jul 31 '15 at 14:07

You are not cleaning up the server after it closes. Meaning you are leaving dead socket resources lying around, which the OS doesn't clean up immediately after the process ends.

You need to call httpd.server_close() in a finally block after your call to httpd.serve_forever(). This call tells the OS to free up any resources that may have been associated with the given server instance.

try:
    httpd.serve_forever()
finally:
    httpd.server_close()

Python's SimpleHTTPServer started in a thread won't close the port

Update

3 Answers3