Related
I am sketching the architecture for a set of programs that share various interrelated objects stored in a database. I want one of the programs to act as a service which provides a higher level interface for operations on these objects, and the other programs to access the objects through that service.
I am currently aiming for Python and the Django framework as the technologies to implement that service with. I'm pretty sure I figure how to daemonize the Python program in Linux. However, it is an optional spec item that the system should support Windows. I have little experience with Windows programming and no experience at all with Windows services.
Is it possible to run a Python programs as a Windows service (i. e. run it automatically without user login)? I won't necessarily have to implement this part, but I need a rough idea how it would be done in order to decide whether to design along these lines.
Edit: Thanks for all the answers so far, they are quite comprehensive. I would like to know one more thing: How is Windows aware of my service? Can I manage it with the native Windows utilities? What is the equivalent of putting a start/stop script in /etc/init.d?
Yes you can. I do it using the pythoncom libraries that come included with ActivePython or can be installed with pywin32 (Python for Windows extensions).
This is a basic skeleton for a simple service:
import win32serviceutil
import win32service
import win32event
import servicemanager
import socket
class AppServerSvc (win32serviceutil.ServiceFramework):
_svc_name_ = "TestService"
_svc_display_name_ = "Test Service"
def __init__(self,args):
win32serviceutil.ServiceFramework.__init__(self,args)
self.hWaitStop = win32event.CreateEvent(None,0,0,None)
socket.setdefaulttimeout(60)
def SvcStop(self):
self.ReportServiceStatus(win32service.SERVICE_STOP_PENDING)
win32event.SetEvent(self.hWaitStop)
def SvcDoRun(self):
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE,
servicemanager.PYS_SERVICE_STARTED,
(self._svc_name_,''))
self.main()
def main(self):
pass
if __name__ == '__main__':
win32serviceutil.HandleCommandLine(AppServerSvc)
Your code would go in the main() method—usually with some kind of infinite loop that might be interrupted by checking a flag, which you set in the SvcStop method
The simplest way is to use the: NSSM - the Non-Sucking Service Manager. Just download and unzip to a location of your choosing. It's a self-contained utility, around 300KB (much less than installing the entire pywin32 suite just for this purpose) and no "installation" is needed. The zip contains a 64-bit and a 32-bit version of the utility. Either should work well on current systems (you can use the 32-bit version to manage services on 64-bit systems).
GUI approach
1 - install the python program as a service. Open a Win prompt as admin
c:\>nssm.exe install WinService
2 - On NSSM´s GUI console:
path: C:\Python27\Python27.exe
Startup directory: C:\Python27
Arguments: c:\WinService.py
3 - check the created services on services.msc
Scripting approach (no GUI)
This is handy if your service should be part of an automated, non-interactive procedure, that may be beyond your control, such as a batch or installer script. It is assumed that the commands are executed with administrative privileges.
For convenience the commands are described here by simply referring to the utility as nssm.exe. It is advisable, however, to refer to it more explicitly in scripting with its full path c:\path\to\nssm.exe, since it's a self-contained executable that may be located in a private path that the system is not aware of.
1. Install the service
You must specify a name for the service, the path to the proper Python executable, and the path to the script:
nssm.exe install ProjectService "c:\path\to\python.exe" "c:\path\to\project\app\main.py"
More explicitly:
nssm.exe install ProjectService
nssm.exe set ProjectService Application "c:\path\to\python.exe"
nssm.exe set ProjectService AppParameters "c:\path\to\project\app\main.py"
Alternatively you may want your Python app to be started as a Python module. One easy approach is to tell nssm that it needs to change to the proper starting directory, as you would do yourself when launching from a command shell:
nssm.exe install ProjectService "c:\path\to\python.exe" "-m app.main"
nssm.exe set ProjectService AppDirectory "c:\path\to\project"
This approach works well with virtual environments and self-contained (embedded) Python installs. Just make sure to have properly resolved any path issues in those environments with the usual methods. nssm has a way to set environment variables (e.g. PYTHONPATH) if needed, and can also launch batch scripts.
2. To start the service
nssm.exe start ProjectService
3. To stop the service
nssm.exe stop ProjectService
4. To remove the service, specify the confirm parameter to skip the interactive confirmation.
nssm.exe remove ProjectService confirm
Although I upvoted the chosen answer a couple of weeks back, in the meantime I struggled a lot more with this topic. It feels like having a special Python installation and using special modules to run a script as a service is simply the wrong way. What about portability and such?
I stumbled across the wonderful Non-sucking Service Manager, which made it really simple and sane to deal with Windows Services. I figured since I could pass options to an installed service, I could just as well select my Python executable and pass my script as an option.
I have not yet tried this solution, but I will do so right now and update this post along the process. I am also interested in using virtualenvs on Windows, so I might come up with a tutorial sooner or later and link to it here.
The simplest way to achieve this is to use native command sc.exe:
sc create PythonApp binPath= "C:\Python34\Python.exe --C:\tmp\pythonscript.py"
References:
https://technet.microsoft.com/en-us/library/cc990289(v=ws.11).aspx
When creating a service with sc.exe how to pass in context parameters?
There are a couple alternatives for installing as a service virtually any Windows executable.
Method 1: Use instsrv and srvany from rktools.exe
For Windows Home Server or Windows Server 2003 (works with WinXP too), the Windows Server 2003 Resource Kit Tools comes with utilities that can be used in tandem for this, called instsrv.exe and srvany.exe. See this Microsoft KB article KB137890 for details on how to use these utils.
For Windows Home Server, there is a great user friendly wrapper for these utilities named aptly "Any Service Installer".
Method 2: Use ServiceInstaller for Windows NT
There is another alternative using ServiceInstaller for Windows NT (download-able here) with python instructions available. Contrary to the name, it works with both Windows 2000 and Windows XP as well. Here are some instructions for how to install a python script as a service.
Installing a Python script
Run ServiceInstaller to create a new
service. (In this example, it is
assumed that python is installed at
c:\python25)
Service Name : PythonTest
Display Name : PythonTest
Startup : Manual (or whatever you like)
Dependencies : (Leave blank or fill to fit your needs)
Executable : c:\python25\python.exe
Arguments : c:\path_to_your_python_script\test.py
Working Directory : c:\path_to_your_python_script
After installing, open the Control
Panel's Services applet, select and
start the PythonTest service.
After my initial answer, I noticed there were closely related Q&A already posted on SO. See also:
Can I run a Python script as a service (in Windows)? How?
How do I make Windows aware of a service I have written in Python?
Step by step explanation how to make it work :
1- First create a python file according to the basic skeleton mentioned above. And save it to a path for example : "c:\PythonFiles\AppServerSvc.py"
import win32serviceutil
import win32service
import win32event
import servicemanager
import socket
class AppServerSvc (win32serviceutil.ServiceFramework):
_svc_name_ = "TestService"
_svc_display_name_ = "Test Service"
def __init__(self,args):
win32serviceutil.ServiceFramework.__init__(self,args)
self.hWaitStop = win32event.CreateEvent(None,0,0,None)
socket.setdefaulttimeout(60)
def SvcStop(self):
self.ReportServiceStatus(win32service.SERVICE_STOP_PENDING)
win32event.SetEvent(self.hWaitStop)
def SvcDoRun(self):
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE,
servicemanager.PYS_SERVICE_STARTED,
(self._svc_name_,''))
self.main()
def main(self):
# Your business logic or call to any class should be here
# this time it creates a text.txt and writes Test Service in a daily manner
f = open('C:\\test.txt', 'a')
rc = None
while rc != win32event.WAIT_OBJECT_0:
f.write('Test Service \n')
f.flush()
# block for 24*60*60 seconds and wait for a stop event
# it is used for a one-day loop
rc = win32event.WaitForSingleObject(self.hWaitStop, 24 * 60 * 60 * 1000)
f.write('shut down \n')
f.close()
if __name__ == '__main__':
win32serviceutil.HandleCommandLine(AppServerSvc)
2 - On this step we should register our service.
Run command prompt as administrator and type as:
sc create TestService binpath= "C:\Python36\Python.exe c:\PythonFiles\AppServerSvc.py" DisplayName= "TestService" start= auto
the first argument of binpath is the path of python.exe
second argument of binpath is the path of your python file that we created already
Don't miss that you should put one space after every "=" sign.
Then if everything is ok, you should see
[SC] CreateService SUCCESS
Now your python service is installed as windows service now. You can see it in Service Manager and registry under :
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\TestService
3- Ok now. You can start your service on service manager.
You can execute every python file that provides this service skeleton.
pysc: Service Control Manager on Python
Example script to run as a service taken from pythonhosted.org:
from xmlrpc.server import SimpleXMLRPCServer
from pysc import event_stop
class TestServer:
def echo(self, msg):
return msg
if __name__ == '__main__':
server = SimpleXMLRPCServer(('127.0.0.1', 9001))
#event_stop
def stop():
server.server_close()
server.register_instance(TestServer())
server.serve_forever()
Create and start service
import os
import sys
from xmlrpc.client import ServerProxy
import pysc
if __name__ == '__main__':
service_name = 'test_xmlrpc_server'
script_path = os.path.join(
os.path.dirname(__file__), 'xmlrpc_server.py'
)
pysc.create(
service_name=service_name,
cmd=[sys.executable, script_path]
)
pysc.start(service_name)
client = ServerProxy('http://127.0.0.1:9001')
print(client.echo('test scm'))
Stop and delete service
import pysc
service_name = 'test_xmlrpc_server'
pysc.stop(service_name)
pysc.delete(service_name)
pip install pysc
nssm in python 3+
(I converted my .py file to .exe with pyinstaller)
nssm:
as said before
run nssm install {ServiceName}
On NSSM´s console:
path: path\to\your\program.exe
Startup directory: path\to\your\ #same as the path but without your program.exe
Arguments: empty
If you don't want to convert your project to .exe
create a .bat file with python {{your python.py file name}}
and set the path to the .bat file
I started hosting as a service with pywin32.
Everything was well but I met the problem that service was not able to start within 30 seconds (default timeout for Windows) on system startup. It was critical for me because Windows startup took place simultaneous on several virtual machines hosted on one physical machine, and IO load was huge.
Error messages were:
Error 1053: The service did not respond to the start or control request in a timely fashion.
Error 7009: Timeout (30000 milliseconds) waiting for the <ServiceName> service to connect.
I fought a lot with pywin, but ended up with using NSSM as it was proposed in this answer. It was very easy to migrate to it.
A complete pywin32 example using loop or subthread
After working on this on and off for a few days, here is the answer I would have wished to find, using pywin32 to keep it nice and self contained.
This is complete working code for one loop-based and one thread-based solution.
It may work on both python 2 and 3, although I've only tested the latest version on 2.7 and Win7. The loop should be good for polling code, and the tread should work with more server-like code. It seems to work nicely with the waitress wsgi server that does not have a standard way to shut down gracefully.
I would also like to note that there seems to be loads of examples out there, like this that are almost useful, but in reality misleading, because they have cut and pasted other examples blindly. I could be wrong. but why create an event if you never wait for it?
That said I still feel I'm on somewhat shaky ground here, especially with regards to how clean the exit from the thread version is, but at least I believe there are nothing misleading here.
To run simply copy the code to a file and follow the instructions.
update:
Use a simple flag to terminate thread. The important bit is that "thread done" prints.
For a more elaborate example exiting from an uncooperative server thread see my post about the waitress wsgi server.
# uncomment mainthread() or mainloop() call below
# run without parameters to see HandleCommandLine options
# install service with "install" and remove with "remove"
# run with "debug" to see print statements
# with "start" and "stop" watch for files to appear
# check Windows EventViever for log messages
import socket
import sys
import threading
import time
from random import randint
from os import path
import servicemanager
import win32event
import win32service
import win32serviceutil
# see http://timgolden.me.uk/pywin32-docs/contents.html for details
def dummytask_once(msg='once'):
fn = path.join(path.dirname(__file__),
'%s_%s.txt' % (msg, randint(1, 10000)))
with open(fn, 'w') as fh:
print(fn)
fh.write('')
def dummytask_loop():
global do_run
while do_run:
dummytask_once(msg='loop')
time.sleep(3)
class MyThread(threading.Thread):
def __init__(self):
threading.Thread.__init__(self)
def run(self):
global do_run
do_run = True
print('thread start\n')
dummytask_loop()
print('thread done\n')
def exit(self):
global do_run
do_run = False
class SMWinservice(win32serviceutil.ServiceFramework):
_svc_name_ = 'PyWinSvc'
_svc_display_name_ = 'Python Windows Service'
_svc_description_ = 'An example of a windows service in Python'
#classmethod
def parse_command_line(cls):
win32serviceutil.HandleCommandLine(cls)
def __init__(self, args):
win32serviceutil.ServiceFramework.__init__(self, args)
self.stopEvt = win32event.CreateEvent(None, 0, 0, None) # create generic event
socket.setdefaulttimeout(60)
def SvcStop(self):
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE,
servicemanager.PYS_SERVICE_STOPPED,
(self._svc_name_, ''))
self.ReportServiceStatus(win32service.SERVICE_STOP_PENDING)
win32event.SetEvent(self.stopEvt) # raise event
def SvcDoRun(self):
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE,
servicemanager.PYS_SERVICE_STARTED,
(self._svc_name_, ''))
# UNCOMMENT ONE OF THESE
# self.mainthread()
# self.mainloop()
# Wait for stopEvt indefinitely after starting thread.
def mainthread(self):
print('main start')
self.server = MyThread()
self.server.start()
print('wait for win32event')
win32event.WaitForSingleObject(self.stopEvt, win32event.INFINITE)
self.server.exit()
print('wait for thread')
self.server.join()
print('main done')
# Wait for stopEvt event in loop.
def mainloop(self):
print('loop start')
rc = None
while rc != win32event.WAIT_OBJECT_0:
dummytask_once()
rc = win32event.WaitForSingleObject(self.stopEvt, 3000)
print('loop done')
if __name__ == '__main__':
SMWinservice.parse_command_line()
This answer is plagiarizer from several sources on StackOverflow - most of them above, but I've forgotten the others - sorry. It's simple and scripts run "as is". For releases you test you script, then copy it to the server and Stop/Start the associated service. And it should work for all scripting languages (Python, Perl, node.js), plus batch scripts such as GitBash, PowerShell, even old DOS bat scripts.
pyGlue is the glue that sits between Windows Services and your script.
'''
A script to create a Windows Service, which, when started, will run an executable with the specified parameters.
Optionally, you can also specify a startup directory
To use this script you MUST define (in class Service)
1. A name for your service (short - preferably no spaces)
2. A display name for your service (the name visibile in Windows Services)
3. A description for your service (long details visible when you inspect the service in Windows Services)
4. The full path of the executable (usually C:/Python38/python.exe or C:WINDOWS/System32/WindowsPowerShell/v1.0/powershell.exe
5. The script which Python or PowerShell will run(or specify None if your executable is standalone - in which case you don't need pyGlue)
6. The startup directory (or specify None)
7. Any parameters for your script (or for your executable if you have no script)
NOTE: This does not make a portable script.
The associated '_svc_name.exe' in the dist folder will only work if the executable,
(and any optional startup directory) actually exist in those locations on the target system
Usage: 'pyGlue.exe [options] install|update|remove|start [...]|stop|restart [...]|debug [...]'
Options for 'install' and 'update' commands only:
--username domain\\username : The Username the service is to run under
--password password : The password for the username
--startup [manual|auto|disabled|delayed] : How the service starts, default = manual
--interactive : Allow the service to interact with the desktop.
--perfmonini file: .ini file to use for registering performance monitor data
--perfmondll file: .dll file to use when querying the service for performance data, default = perfmondata.dll
Options for 'start' and 'stop' commands only:
--wait seconds: Wait for the service to actually start or stop.
If you specify --wait with the 'stop' option, the service and all dependent services will be stopped,
each waiting the specified period.
'''
# Import all the modules that make life easy
import servicemanager
import socket
import sys
import win32event
import win32service
import win32serviceutil
import win32evtlogutil
import os
from logging import Formatter, Handler
import logging
import subprocess
# Define the win32api class
class Service (win32serviceutil.ServiceFramework):
# The following variable are edited by the build.sh script
_svc_name_ = "TestService"
_svc_display_name_ = "Test Service"
_svc_description_ = "Test Running Python Scripts as a Service"
service_exe = 'c:/Python27/python.exe'
service_script = None
service_params = []
service_startDir = None
# Initialize the service
def __init__(self, args):
win32serviceutil.ServiceFramework.__init__(self, args)
self.hWaitStop = win32event.CreateEvent(None, 0, 0, None)
self.configure_logging()
socket.setdefaulttimeout(60)
# Configure logging to the WINDOWS Event logs
def configure_logging(self):
self.formatter = Formatter('%(message)s')
self.handler = logHandler()
self.handler.setFormatter(self.formatter)
self.logger = logging.getLogger()
self.logger.addHandler(self.handler)
self.logger.setLevel(logging.INFO)
# Stop the service
def SvcStop(self):
self.ReportServiceStatus(win32service.SERVICE_STOP_PENDING)
win32event.SetEvent(self.hWaitStop)
# Run the service
def SvcDoRun(self):
self.main()
# This is the service
def main(self):
# Log that we are starting
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE, servicemanager.PYS_SERVICE_STARTED,
(self._svc_name_, ''))
# Fire off the real process that does the real work
logging.info('%s - about to call Popen() to run %s %s %s', self._svc_name_, self.service_exe, self.service_script, self.service_params)
self.process = subprocess.Popen([self.service_exe, self.service_script] + self.service_params, shell=False, cwd=self.service_startDir)
logging.info('%s - started process %d', self._svc_name_, self.process.pid)
# Wait until WINDOWS kills us - retrigger the wait for stop every 60 seconds
rc = None
while rc != win32event.WAIT_OBJECT_0:
rc = win32event.WaitForSingleObject(self.hWaitStop, (1 * 60 * 1000))
# Shut down the real process and exit
logging.info('%s - is terminating process %d', self._svc_name_, self.process.pid)
self.process.terminate()
logging.info('%s - is exiting', self._svc_name_)
class logHandler(Handler):
'''
Emit a log record to the WINDOWS Event log
'''
def emit(self, record):
servicemanager.LogInfoMsg(record.getMessage())
# The main code
if __name__ == '__main__':
'''
Create a Windows Service, which, when started, will run an executable with the specified parameters.
'''
# Check that configuration contains valid values just in case this service has accidentally
# been moved to a server where things are in different places
if not os.path.isfile(Service.service_exe):
print('Executable file({!s}) does not exist'.format(Service.service_exe), file=sys.stderr)
sys.exit(0)
if not os.access(Service.service_exe, os.X_OK):
print('Executable file({!s}) is not executable'.format(Service.service_exe), file=sys.stderr)
sys.exit(0)
# Check that any optional startup directory exists
if (Service.service_startDir is not None) and (not os.path.isdir(Service.service_startDir)):
print('Start up directory({!s}) does not exist'.format(Service.service_startDir), file=sys.stderr)
sys.exit(0)
if len(sys.argv) == 1:
servicemanager.Initialize()
servicemanager.PrepareToHostSingle(Service)
servicemanager.StartServiceCtrlDispatcher()
else:
# install/update/remove/start/stop/restart or debug the service
# One of those command line options must be specified
win32serviceutil.HandleCommandLine(Service)
Now there's a bit of editing and you don't want all your services called 'pyGlue'. So there's a script (build.sh) to plug in the bits and create a customized 'pyGlue' and create an '.exe'. It is this '.exe' which gets installed as a Windows Service. Once installed you can set it to run automatically.
#!/bin/sh
# This script build a Windows Service that will install/start/stop/remove a service that runs a script
# That is, executes Python to run a Python script, or PowerShell to run a PowerShell script, etc
if [ $# -lt 6 ]; then
echo "Usage: build.sh Name Display Description Executable Script StartupDir [Params]..."
exit 0
fi
name=$1
display=$2
desc=$3
exe=$4
script=$5
startDir=$6
shift; shift; shift; shift; shift; shift
params=
while [ $# -gt 0 ]; do
if [ "${params}" != "" ]; then
params="${params}, "
fi
params="${params}'$1'"
shift
done
cat pyGlue.py | sed -e "s/pyGlue/${name}/g" | \
sed -e "/_svc_name_ =/s?=.*?= '${name}'?" | \
sed -e "/_svc_display_name_ =/s?=.*?= '${display}'?" | \
sed -e "/_svc_description_ =/s?=.*?= '${desc}'?" | \
sed -e "/service_exe =/s?=.*?= '$exe'?" | \
sed -e "/service_script =/s?=.*?= '$script'?" | \
sed -e "/service_params =/s?=.*?= [${params}]?" | \
sed -e "/service_startDir =/s?=.*?= '${startDir}'?" > ${name}.py
cxfreeze ${name}.py --include-modules=win32timezone
Installation - copy the '.exe' the server and the script to the specified folder. Run the '.exe', as Administrator, with the 'install' option. Open Windows Services, as Adminstrator, and start you service. For upgrade, just copy the new version of the script and Stop/Start the service.
Now every server is different - different installations of Python, different folder structures. I maintain a folder for every server, with a copy of pyGlue.py and build.sh. And I create a 'serverBuild.sh' script for rebuilding all the service on that server.
# A script to build all the script based Services on this PC
sh build.sh AutoCode 'AutoCode Medical Documents' 'Autocode Medical Documents to SNOMED_CT and AIHW codes' C:/Python38/python.exe autocode.py C:/Users/russell/Documents/autocoding -S -T
The accepted answer using win32serviceutil works but is complicated and makes debugging and changes harder. It is far easier to use NSSM (the Non-Sucking Service Manager). You write and comfortably debug a normal python program and when it finally works you use NSSM to install it as a service in less than a minute:
From an elevated (admin) command prompt you run nssm.exe install NameOfYourService and you fill-in these options:
path: (the path to python.exe e.g. C:\Python27\Python.exe)
Arguments: (the path to your python script, e.g. c:\path\to\program.py)
By the way, if your program prints useful messages that you want to keep in a log file NSSM can also handle this and a lot more for you.
This doesn't answer the original question, but might help other people that want to automatically start a Python script at Windows startup:
Have a look at the Windows Task Scheduler instead, it is way easier if you just want to start a script after boot without all the service functionality of a Windows Service.
Create a new task, select "At startup" as trigger, "Start program" as action with "C:\Python39\python.exe" as the program (or wherever your python.exe is) and the full path to your script ("C:...\my_dir\xyz.py") as argument (you can use " if the path contains spaces).
You can also select the path of your script (without the .py file, e.g. "C:...\my_dir") for "start in" if you use relative paths in your script, e.g. for logging.
https://www.chrisumbel.com/article/windows_services_in_python
Follow up the PySvc.py
changing the dll folder
I know this is old but I was stuck on this forever. For me, this specific problem was solved by copying this file - pywintypes36.dll
From -> Python36\Lib\site-packages\pywin32_system32
To -> Python36\Lib\site-packages\win32
setx /M PATH "%PATH%;C:\Users\user\AppData\Local\Programs\Python\Python38-32;C:\Users\user\AppData\Local\Programs\Python\Python38-32\Scripts;C:\Users\user\AppData\Local\Programs\Python\Python38-32\Lib\site-packages\pywin32_system32;C:\Users\user\AppData\Local\Programs\Python\Python38-32\Lib\site-packages\win32
changing the path to python folder by
cd C:\Users\user\AppData\Local\Programs\Python\Python38-32
NET START PySvc
NET STOP PySvc
I'm relatively new to python so please forgive early level understanding!
I am working to create a kind of flag file. Its job is to monitor a Python executable, the flag file is constantly running and prints "Start" when the executable started, "Running" while it runs and "Stop" when its stopped or crashed, if a crash occurs i want it to be able to restart the script. so far i have this down for the Restart:
from subprocess import run
from time import sleep
# Path and name to the script you are trying to start
file_path = "py"
restart_timer = 2
def start_script():
try:
# Make sure 'python' command is available
run("python "+file_path, check=True)
except:
# Script crashed, lets restart it!
handle_crash()
def handle_crash():
sleep(restart_timer) # Restarts the script after 2 seconds
start_script()
start_script()
how can i implement this along with a flag file?
Not sure what you mean with "flag", but this minimally achieves what you want.
Main file main.py:
import subprocess
import sys
from time import sleep
restart_timer = 2
file_path = 'sub.py' # file name of the other process
def start():
try:
# sys.executable -> same python executable
subprocess.run([sys.executable, file_path], check=True)
except subprocess.CalledProcessError:
sleep(restart_timer)
return True
else:
return False
def main():
print("starting...")
monitor = True
while monitor:
monitor = start()
if __name__ == '__main__':
main()
Then the process that gets spawned, called sub.py:
from time import sleep
sleep(1)
print("doing stuff...")
# comment out to see change
raise ValueError("sub.py is throwing error...")
Put those files into the same directory and run it with python main.py
You can comment out the throwing of the random error to see the main script terminate normally.
On a larger note, this example is not saying it is a good way to achieve the quality you need...
I have a Django app in which the users can create reports at various places throughout the webapp. As a little nifty feature I would like to make a "send as PDF" feature for these reports, and I'm almost there.
What I do is that before the report is returned as a HttpResponse through Django I send the raw HTML content through a small PySide/QT snippet (as seen below).
The problem is that I can't get the QApplication to quit.
I have tried with the standard QCoreApplication.exit() but with no luck. If I try to convert a new report right after the first one, the console says a "QApplication instance already exists".
I am using Django 1.2.5, Python 2.7, QT 4.8 and PySide 1.1 on OS X 10.7.3 (for testing).
Code:
def makepdf(response,filename):
try:
app = QApplication(sys.argv)
except:
app = QCoreApplication.instance()
web = QWebView()
stylelink = "%s%s" % (media_root,'images/css/reportGenerator.css')
web.settings().setUserStyleSheetUrl(QUrl.fromLocalFile(stylelink))
web.setHtml(response)
printer = QPrinter()
printer.setPageSize(QPrinter.A4)
printer.setOutputFormat(QPrinter.PdfFormat)
printer.setOutputFileName(filename)
def convertToPdf():
web.print_(printer)
#webresponse = web.close()
#QObject.disconnect()
#print QCoreApplication.instance().children()[0].interrupt()
#qthread = QCoreApplication.instance().thread()#.cleanup()
#qthread.exit()
QCoreApplication.exit()
QObject.connect(web, SIGNAL("loadFinished(bool)"), convertToPdf())
Code comments:
Currently I have made a try/except clause to be able to keep the code running, by using the current QApplication instance (to avoid the 'instance exists error'), but this just doesn't seem right?. I mean, having a QAppliation running for the duration of my Apache server (which will be the case in production) seems a bit off. Shouldn't it be possible to quit it after the PDF conversion is done?
Other than the QCoreApplication.exit() I have tried to use the sys.exit(app.exec_()) method, as is often seen in examples and snippets. This however just causes an error and makes Python crash and the code works perfectly without this. (well it creates the PDF, but won't quit).
All the lines commented out are the previous attempts I have made to quit the QApplication.
In short: I don't know what it is, but it just won't quit. Can anyone suggest why?
Update: After the latest answer I have edited the code to respond to the input. This is how the final part looks now:
def convertToPdf():
web.print_(printer)
app.exit()
web.loadFinished.connect(convertToPdf)
app.exec_()
I do, however, still get an error:
2012-04-30 00:16:10.791 Python[21241:1803] * Assertion failure in +[NSUndoManager _endTopLevelGroupings], /SourceCache/Foundation/Foundation-833.24/Misc.subproj/NSUndoManager.m:324
Qt has caught an exception thrown from an event handler. Throwing exceptions from an event handler is not supported in Qt. You must reimplement QApplication::notify() and catch all exceptions there.
This error only occurs when I implement app.exec_(). However without app.exec_() it is just the normal issue again, with no quitting.
Any ideas?
Update 2: this is the newest code, fixed in accordance with matas latest suggestion:
app = QApplication(sys.argv)
web = QWebView()
printer = QPrinter()
printer.setPageSize(QPrinter.A4)
printer.setOutputFormat(QPrinter.PdfFormat)
printer.setOutputFileName(filename)
def convertToPdf():
web.print_(printer)
app.exit()
web.loadFinished.connect(convertToPdf)
web.setHtml(response)
I still have the same problem however.
what i can say is this:
QObject.connect(web, SIGNAL("loadFinished(bool)"), convertToPdf())
here you call convertToPdf(), if you want to connect the signal, omit the parenthesis!
you could also use this much clearer syntax:
web.loadFinished.connect(convertToPdf)
you may also want to add a parameter to convertToPdf, as it is called with a boolean indicating wheather loading was successful or not.
And using app.exit() should be enough.
oh, and when you use Gui-Components you need to use a QApplication. A QCoreApplication won't do!
edit: it's important to call web.setHtml after you've connected the loadFinished signal! otherwise if the loading already is finished, your function will never be executed!
edit: this works without any problem for me:
#!/usr/bin/env python
from PySide.QtGui import *
from PySide.QtWebKit import *
import sys
from subprocess import Popen, PIPE
def createPdf(html, filename):
app = QApplication(sys.argv)
web = QWebView()
printer = QPrinter()
printer.setPageSize(QPrinter.A4)
printer.setOutputFormat(QPrinter.PdfFormat)
printer.setOutputFileName(filename)
def convertToPdf():
web.print_(printer)
app.exit()
app.deleteLater()
web.loadFinished.connect(convertToPdf)
web.setHtml(html)
app.processEvents()
def createPdfInSubprocess(html, filename):
p = Popen(["python", __file__, filename],
stdin=PIPE, stdout=PIPE, stderr=PIPE)
out, err = p.communicate(html)
return (p.returncode, out, err)
if __name__ == '__main__':
if len(sys.argv) > 1:
# read html from stdin, filename from cmdline
html = "\n".join(sys.stdin.readlines())
createPdf(html, sys.argv[1])
# print("done")
else:
# test if it's working
ret = createPdfInSubprocess(
"<html><h1>test</h1>it's working...</html>", "test.pdf")
print(ret)
Even without all the calls to app.exit(), app.deleteLater(), app.processEvents(), it still works... but can't hurt to have that.
One more important thing: QApplications have to be created from the main thread! So if it's running inside a django app it's probably not going to work, that's why i added the subprocess stuff...
PySide is designed such that there is only ever one instance of QCoreApplication within a process. The best reference to this (possibly undocumented fact) that I could find was http://bugs.pyside.org/show_bug.cgi?id=855.
Basically, there may be dangling references to the qapplication that prevent it from being reaped by the garbage collector, so even if you tell the application to exit and you delete your reference, there may still be other references lying around.
I've developed a Python application that runs both in the GUI mode and the console mode. If any arguments are specified, it runs in a console mode else it runs in the GUI mode.
I've managed to freeze this using cx_Freeze. I had some problems hiding the black console window that would pop up with wxPython and so I modified my setup.py script like this:
import sys
from cx_Freeze import setup, Executable
base = None
if sys.platform == "win32":
base = "Win32GUI"
setup(
name = "simple_PyQt4",
version = "0.1",
description = "Sample cx_Freeze PyQt4 script",
executables = [Executable("PyQt4app.py", base = base)])
This works fine but now when I try to open up my console and run the executable from there, it doesn't output anything. I don't get any errors or messages so it seems that cx_Feeze is redirecting the stdout somewhere else.
Is is possible to get it to work with both mode? Nothing similar to this seems to be documented anywhere. :(
Thanks in advance.
Mridang
I found this bit on this page:
Tip for the console-less version: If
you try to print anything, you will
get a nasty error window, because
stdout and stderr do not exist (and
the cx_freeze Win32gui.exe stub will
display an error Window). This is a
pain when you want your program to be
able to run in GUI mode and
command-line mode. To safely disable
console output, do as follows at the
beginning of your program:
try:
sys.stdout.write("\n")
sys.stdout.flush()
except IOError:
class dummyStream:
''' dummyStream behaves like a stream but does nothing. '''
def __init__(self): pass
def write(self,data): pass
def read(self,data): pass
def flush(self): pass
def close(self): pass
# and now redirect all default streams to this dummyStream:
sys.stdout = dummyStream()
sys.stderr = dummyStream()
sys.stdin = dummyStream()
sys.__stdout__ = dummyStream()
sys.__stderr__ = dummyStream()
sys.__stdin__ = dummyStream()
This way, if the program starts in
console-less mode, it will work even
if the code contains print statements.
And if run in command-line mode, it
will print out as usual. (This is
basically what I did in webGobbler,
too.)
Raymond Chen has written about this. In short, it's not possible directly under Windows but there are some workarounds.
I'd suggest shipping two executables - a CLI and GUI one.
I am sketching the architecture for a set of programs that share various interrelated objects stored in a database. I want one of the programs to act as a service which provides a higher level interface for operations on these objects, and the other programs to access the objects through that service.
I am currently aiming for Python and the Django framework as the technologies to implement that service with. I'm pretty sure I figure how to daemonize the Python program in Linux. However, it is an optional spec item that the system should support Windows. I have little experience with Windows programming and no experience at all with Windows services.
Is it possible to run a Python programs as a Windows service (i. e. run it automatically without user login)? I won't necessarily have to implement this part, but I need a rough idea how it would be done in order to decide whether to design along these lines.
Edit: Thanks for all the answers so far, they are quite comprehensive. I would like to know one more thing: How is Windows aware of my service? Can I manage it with the native Windows utilities? What is the equivalent of putting a start/stop script in /etc/init.d?
Yes you can. I do it using the pythoncom libraries that come included with ActivePython or can be installed with pywin32 (Python for Windows extensions).
This is a basic skeleton for a simple service:
import win32serviceutil
import win32service
import win32event
import servicemanager
import socket
class AppServerSvc (win32serviceutil.ServiceFramework):
_svc_name_ = "TestService"
_svc_display_name_ = "Test Service"
def __init__(self,args):
win32serviceutil.ServiceFramework.__init__(self,args)
self.hWaitStop = win32event.CreateEvent(None,0,0,None)
socket.setdefaulttimeout(60)
def SvcStop(self):
self.ReportServiceStatus(win32service.SERVICE_STOP_PENDING)
win32event.SetEvent(self.hWaitStop)
def SvcDoRun(self):
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE,
servicemanager.PYS_SERVICE_STARTED,
(self._svc_name_,''))
self.main()
def main(self):
pass
if __name__ == '__main__':
win32serviceutil.HandleCommandLine(AppServerSvc)
Your code would go in the main() method—usually with some kind of infinite loop that might be interrupted by checking a flag, which you set in the SvcStop method
The simplest way is to use the: NSSM - the Non-Sucking Service Manager. Just download and unzip to a location of your choosing. It's a self-contained utility, around 300KB (much less than installing the entire pywin32 suite just for this purpose) and no "installation" is needed. The zip contains a 64-bit and a 32-bit version of the utility. Either should work well on current systems (you can use the 32-bit version to manage services on 64-bit systems).
GUI approach
1 - install the python program as a service. Open a Win prompt as admin
c:\>nssm.exe install WinService
2 - On NSSM´s GUI console:
path: C:\Python27\Python27.exe
Startup directory: C:\Python27
Arguments: c:\WinService.py
3 - check the created services on services.msc
Scripting approach (no GUI)
This is handy if your service should be part of an automated, non-interactive procedure, that may be beyond your control, such as a batch or installer script. It is assumed that the commands are executed with administrative privileges.
For convenience the commands are described here by simply referring to the utility as nssm.exe. It is advisable, however, to refer to it more explicitly in scripting with its full path c:\path\to\nssm.exe, since it's a self-contained executable that may be located in a private path that the system is not aware of.
1. Install the service
You must specify a name for the service, the path to the proper Python executable, and the path to the script:
nssm.exe install ProjectService "c:\path\to\python.exe" "c:\path\to\project\app\main.py"
More explicitly:
nssm.exe install ProjectService
nssm.exe set ProjectService Application "c:\path\to\python.exe"
nssm.exe set ProjectService AppParameters "c:\path\to\project\app\main.py"
Alternatively you may want your Python app to be started as a Python module. One easy approach is to tell nssm that it needs to change to the proper starting directory, as you would do yourself when launching from a command shell:
nssm.exe install ProjectService "c:\path\to\python.exe" "-m app.main"
nssm.exe set ProjectService AppDirectory "c:\path\to\project"
This approach works well with virtual environments and self-contained (embedded) Python installs. Just make sure to have properly resolved any path issues in those environments with the usual methods. nssm has a way to set environment variables (e.g. PYTHONPATH) if needed, and can also launch batch scripts.
2. To start the service
nssm.exe start ProjectService
3. To stop the service
nssm.exe stop ProjectService
4. To remove the service, specify the confirm parameter to skip the interactive confirmation.
nssm.exe remove ProjectService confirm
Although I upvoted the chosen answer a couple of weeks back, in the meantime I struggled a lot more with this topic. It feels like having a special Python installation and using special modules to run a script as a service is simply the wrong way. What about portability and such?
I stumbled across the wonderful Non-sucking Service Manager, which made it really simple and sane to deal with Windows Services. I figured since I could pass options to an installed service, I could just as well select my Python executable and pass my script as an option.
I have not yet tried this solution, but I will do so right now and update this post along the process. I am also interested in using virtualenvs on Windows, so I might come up with a tutorial sooner or later and link to it here.
The simplest way to achieve this is to use native command sc.exe:
sc create PythonApp binPath= "C:\Python34\Python.exe --C:\tmp\pythonscript.py"
References:
https://technet.microsoft.com/en-us/library/cc990289(v=ws.11).aspx
When creating a service with sc.exe how to pass in context parameters?
There are a couple alternatives for installing as a service virtually any Windows executable.
Method 1: Use instsrv and srvany from rktools.exe
For Windows Home Server or Windows Server 2003 (works with WinXP too), the Windows Server 2003 Resource Kit Tools comes with utilities that can be used in tandem for this, called instsrv.exe and srvany.exe. See this Microsoft KB article KB137890 for details on how to use these utils.
For Windows Home Server, there is a great user friendly wrapper for these utilities named aptly "Any Service Installer".
Method 2: Use ServiceInstaller for Windows NT
There is another alternative using ServiceInstaller for Windows NT (download-able here) with python instructions available. Contrary to the name, it works with both Windows 2000 and Windows XP as well. Here are some instructions for how to install a python script as a service.
Installing a Python script
Run ServiceInstaller to create a new
service. (In this example, it is
assumed that python is installed at
c:\python25)
Service Name : PythonTest
Display Name : PythonTest
Startup : Manual (or whatever you like)
Dependencies : (Leave blank or fill to fit your needs)
Executable : c:\python25\python.exe
Arguments : c:\path_to_your_python_script\test.py
Working Directory : c:\path_to_your_python_script
After installing, open the Control
Panel's Services applet, select and
start the PythonTest service.
After my initial answer, I noticed there were closely related Q&A already posted on SO. See also:
Can I run a Python script as a service (in Windows)? How?
How do I make Windows aware of a service I have written in Python?
Step by step explanation how to make it work :
1- First create a python file according to the basic skeleton mentioned above. And save it to a path for example : "c:\PythonFiles\AppServerSvc.py"
import win32serviceutil
import win32service
import win32event
import servicemanager
import socket
class AppServerSvc (win32serviceutil.ServiceFramework):
_svc_name_ = "TestService"
_svc_display_name_ = "Test Service"
def __init__(self,args):
win32serviceutil.ServiceFramework.__init__(self,args)
self.hWaitStop = win32event.CreateEvent(None,0,0,None)
socket.setdefaulttimeout(60)
def SvcStop(self):
self.ReportServiceStatus(win32service.SERVICE_STOP_PENDING)
win32event.SetEvent(self.hWaitStop)
def SvcDoRun(self):
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE,
servicemanager.PYS_SERVICE_STARTED,
(self._svc_name_,''))
self.main()
def main(self):
# Your business logic or call to any class should be here
# this time it creates a text.txt and writes Test Service in a daily manner
f = open('C:\\test.txt', 'a')
rc = None
while rc != win32event.WAIT_OBJECT_0:
f.write('Test Service \n')
f.flush()
# block for 24*60*60 seconds and wait for a stop event
# it is used for a one-day loop
rc = win32event.WaitForSingleObject(self.hWaitStop, 24 * 60 * 60 * 1000)
f.write('shut down \n')
f.close()
if __name__ == '__main__':
win32serviceutil.HandleCommandLine(AppServerSvc)
2 - On this step we should register our service.
Run command prompt as administrator and type as:
sc create TestService binpath= "C:\Python36\Python.exe c:\PythonFiles\AppServerSvc.py" DisplayName= "TestService" start= auto
the first argument of binpath is the path of python.exe
second argument of binpath is the path of your python file that we created already
Don't miss that you should put one space after every "=" sign.
Then if everything is ok, you should see
[SC] CreateService SUCCESS
Now your python service is installed as windows service now. You can see it in Service Manager and registry under :
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\TestService
3- Ok now. You can start your service on service manager.
You can execute every python file that provides this service skeleton.
pysc: Service Control Manager on Python
Example script to run as a service taken from pythonhosted.org:
from xmlrpc.server import SimpleXMLRPCServer
from pysc import event_stop
class TestServer:
def echo(self, msg):
return msg
if __name__ == '__main__':
server = SimpleXMLRPCServer(('127.0.0.1', 9001))
#event_stop
def stop():
server.server_close()
server.register_instance(TestServer())
server.serve_forever()
Create and start service
import os
import sys
from xmlrpc.client import ServerProxy
import pysc
if __name__ == '__main__':
service_name = 'test_xmlrpc_server'
script_path = os.path.join(
os.path.dirname(__file__), 'xmlrpc_server.py'
)
pysc.create(
service_name=service_name,
cmd=[sys.executable, script_path]
)
pysc.start(service_name)
client = ServerProxy('http://127.0.0.1:9001')
print(client.echo('test scm'))
Stop and delete service
import pysc
service_name = 'test_xmlrpc_server'
pysc.stop(service_name)
pysc.delete(service_name)
pip install pysc
nssm in python 3+
(I converted my .py file to .exe with pyinstaller)
nssm:
as said before
run nssm install {ServiceName}
On NSSM´s console:
path: path\to\your\program.exe
Startup directory: path\to\your\ #same as the path but without your program.exe
Arguments: empty
If you don't want to convert your project to .exe
create a .bat file with python {{your python.py file name}}
and set the path to the .bat file
I started hosting as a service with pywin32.
Everything was well but I met the problem that service was not able to start within 30 seconds (default timeout for Windows) on system startup. It was critical for me because Windows startup took place simultaneous on several virtual machines hosted on one physical machine, and IO load was huge.
Error messages were:
Error 1053: The service did not respond to the start or control request in a timely fashion.
Error 7009: Timeout (30000 milliseconds) waiting for the <ServiceName> service to connect.
I fought a lot with pywin, but ended up with using NSSM as it was proposed in this answer. It was very easy to migrate to it.
A complete pywin32 example using loop or subthread
After working on this on and off for a few days, here is the answer I would have wished to find, using pywin32 to keep it nice and self contained.
This is complete working code for one loop-based and one thread-based solution.
It may work on both python 2 and 3, although I've only tested the latest version on 2.7 and Win7. The loop should be good for polling code, and the tread should work with more server-like code. It seems to work nicely with the waitress wsgi server that does not have a standard way to shut down gracefully.
I would also like to note that there seems to be loads of examples out there, like this that are almost useful, but in reality misleading, because they have cut and pasted other examples blindly. I could be wrong. but why create an event if you never wait for it?
That said I still feel I'm on somewhat shaky ground here, especially with regards to how clean the exit from the thread version is, but at least I believe there are nothing misleading here.
To run simply copy the code to a file and follow the instructions.
update:
Use a simple flag to terminate thread. The important bit is that "thread done" prints.
For a more elaborate example exiting from an uncooperative server thread see my post about the waitress wsgi server.
# uncomment mainthread() or mainloop() call below
# run without parameters to see HandleCommandLine options
# install service with "install" and remove with "remove"
# run with "debug" to see print statements
# with "start" and "stop" watch for files to appear
# check Windows EventViever for log messages
import socket
import sys
import threading
import time
from random import randint
from os import path
import servicemanager
import win32event
import win32service
import win32serviceutil
# see http://timgolden.me.uk/pywin32-docs/contents.html for details
def dummytask_once(msg='once'):
fn = path.join(path.dirname(__file__),
'%s_%s.txt' % (msg, randint(1, 10000)))
with open(fn, 'w') as fh:
print(fn)
fh.write('')
def dummytask_loop():
global do_run
while do_run:
dummytask_once(msg='loop')
time.sleep(3)
class MyThread(threading.Thread):
def __init__(self):
threading.Thread.__init__(self)
def run(self):
global do_run
do_run = True
print('thread start\n')
dummytask_loop()
print('thread done\n')
def exit(self):
global do_run
do_run = False
class SMWinservice(win32serviceutil.ServiceFramework):
_svc_name_ = 'PyWinSvc'
_svc_display_name_ = 'Python Windows Service'
_svc_description_ = 'An example of a windows service in Python'
#classmethod
def parse_command_line(cls):
win32serviceutil.HandleCommandLine(cls)
def __init__(self, args):
win32serviceutil.ServiceFramework.__init__(self, args)
self.stopEvt = win32event.CreateEvent(None, 0, 0, None) # create generic event
socket.setdefaulttimeout(60)
def SvcStop(self):
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE,
servicemanager.PYS_SERVICE_STOPPED,
(self._svc_name_, ''))
self.ReportServiceStatus(win32service.SERVICE_STOP_PENDING)
win32event.SetEvent(self.stopEvt) # raise event
def SvcDoRun(self):
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE,
servicemanager.PYS_SERVICE_STARTED,
(self._svc_name_, ''))
# UNCOMMENT ONE OF THESE
# self.mainthread()
# self.mainloop()
# Wait for stopEvt indefinitely after starting thread.
def mainthread(self):
print('main start')
self.server = MyThread()
self.server.start()
print('wait for win32event')
win32event.WaitForSingleObject(self.stopEvt, win32event.INFINITE)
self.server.exit()
print('wait for thread')
self.server.join()
print('main done')
# Wait for stopEvt event in loop.
def mainloop(self):
print('loop start')
rc = None
while rc != win32event.WAIT_OBJECT_0:
dummytask_once()
rc = win32event.WaitForSingleObject(self.stopEvt, 3000)
print('loop done')
if __name__ == '__main__':
SMWinservice.parse_command_line()
This answer is plagiarizer from several sources on StackOverflow - most of them above, but I've forgotten the others - sorry. It's simple and scripts run "as is". For releases you test you script, then copy it to the server and Stop/Start the associated service. And it should work for all scripting languages (Python, Perl, node.js), plus batch scripts such as GitBash, PowerShell, even old DOS bat scripts.
pyGlue is the glue that sits between Windows Services and your script.
'''
A script to create a Windows Service, which, when started, will run an executable with the specified parameters.
Optionally, you can also specify a startup directory
To use this script you MUST define (in class Service)
1. A name for your service (short - preferably no spaces)
2. A display name for your service (the name visibile in Windows Services)
3. A description for your service (long details visible when you inspect the service in Windows Services)
4. The full path of the executable (usually C:/Python38/python.exe or C:WINDOWS/System32/WindowsPowerShell/v1.0/powershell.exe
5. The script which Python or PowerShell will run(or specify None if your executable is standalone - in which case you don't need pyGlue)
6. The startup directory (or specify None)
7. Any parameters for your script (or for your executable if you have no script)
NOTE: This does not make a portable script.
The associated '_svc_name.exe' in the dist folder will only work if the executable,
(and any optional startup directory) actually exist in those locations on the target system
Usage: 'pyGlue.exe [options] install|update|remove|start [...]|stop|restart [...]|debug [...]'
Options for 'install' and 'update' commands only:
--username domain\\username : The Username the service is to run under
--password password : The password for the username
--startup [manual|auto|disabled|delayed] : How the service starts, default = manual
--interactive : Allow the service to interact with the desktop.
--perfmonini file: .ini file to use for registering performance monitor data
--perfmondll file: .dll file to use when querying the service for performance data, default = perfmondata.dll
Options for 'start' and 'stop' commands only:
--wait seconds: Wait for the service to actually start or stop.
If you specify --wait with the 'stop' option, the service and all dependent services will be stopped,
each waiting the specified period.
'''
# Import all the modules that make life easy
import servicemanager
import socket
import sys
import win32event
import win32service
import win32serviceutil
import win32evtlogutil
import os
from logging import Formatter, Handler
import logging
import subprocess
# Define the win32api class
class Service (win32serviceutil.ServiceFramework):
# The following variable are edited by the build.sh script
_svc_name_ = "TestService"
_svc_display_name_ = "Test Service"
_svc_description_ = "Test Running Python Scripts as a Service"
service_exe = 'c:/Python27/python.exe'
service_script = None
service_params = []
service_startDir = None
# Initialize the service
def __init__(self, args):
win32serviceutil.ServiceFramework.__init__(self, args)
self.hWaitStop = win32event.CreateEvent(None, 0, 0, None)
self.configure_logging()
socket.setdefaulttimeout(60)
# Configure logging to the WINDOWS Event logs
def configure_logging(self):
self.formatter = Formatter('%(message)s')
self.handler = logHandler()
self.handler.setFormatter(self.formatter)
self.logger = logging.getLogger()
self.logger.addHandler(self.handler)
self.logger.setLevel(logging.INFO)
# Stop the service
def SvcStop(self):
self.ReportServiceStatus(win32service.SERVICE_STOP_PENDING)
win32event.SetEvent(self.hWaitStop)
# Run the service
def SvcDoRun(self):
self.main()
# This is the service
def main(self):
# Log that we are starting
servicemanager.LogMsg(servicemanager.EVENTLOG_INFORMATION_TYPE, servicemanager.PYS_SERVICE_STARTED,
(self._svc_name_, ''))
# Fire off the real process that does the real work
logging.info('%s - about to call Popen() to run %s %s %s', self._svc_name_, self.service_exe, self.service_script, self.service_params)
self.process = subprocess.Popen([self.service_exe, self.service_script] + self.service_params, shell=False, cwd=self.service_startDir)
logging.info('%s - started process %d', self._svc_name_, self.process.pid)
# Wait until WINDOWS kills us - retrigger the wait for stop every 60 seconds
rc = None
while rc != win32event.WAIT_OBJECT_0:
rc = win32event.WaitForSingleObject(self.hWaitStop, (1 * 60 * 1000))
# Shut down the real process and exit
logging.info('%s - is terminating process %d', self._svc_name_, self.process.pid)
self.process.terminate()
logging.info('%s - is exiting', self._svc_name_)
class logHandler(Handler):
'''
Emit a log record to the WINDOWS Event log
'''
def emit(self, record):
servicemanager.LogInfoMsg(record.getMessage())
# The main code
if __name__ == '__main__':
'''
Create a Windows Service, which, when started, will run an executable with the specified parameters.
'''
# Check that configuration contains valid values just in case this service has accidentally
# been moved to a server where things are in different places
if not os.path.isfile(Service.service_exe):
print('Executable file({!s}) does not exist'.format(Service.service_exe), file=sys.stderr)
sys.exit(0)
if not os.access(Service.service_exe, os.X_OK):
print('Executable file({!s}) is not executable'.format(Service.service_exe), file=sys.stderr)
sys.exit(0)
# Check that any optional startup directory exists
if (Service.service_startDir is not None) and (not os.path.isdir(Service.service_startDir)):
print('Start up directory({!s}) does not exist'.format(Service.service_startDir), file=sys.stderr)
sys.exit(0)
if len(sys.argv) == 1:
servicemanager.Initialize()
servicemanager.PrepareToHostSingle(Service)
servicemanager.StartServiceCtrlDispatcher()
else:
# install/update/remove/start/stop/restart or debug the service
# One of those command line options must be specified
win32serviceutil.HandleCommandLine(Service)
Now there's a bit of editing and you don't want all your services called 'pyGlue'. So there's a script (build.sh) to plug in the bits and create a customized 'pyGlue' and create an '.exe'. It is this '.exe' which gets installed as a Windows Service. Once installed you can set it to run automatically.
#!/bin/sh
# This script build a Windows Service that will install/start/stop/remove a service that runs a script
# That is, executes Python to run a Python script, or PowerShell to run a PowerShell script, etc
if [ $# -lt 6 ]; then
echo "Usage: build.sh Name Display Description Executable Script StartupDir [Params]..."
exit 0
fi
name=$1
display=$2
desc=$3
exe=$4
script=$5
startDir=$6
shift; shift; shift; shift; shift; shift
params=
while [ $# -gt 0 ]; do
if [ "${params}" != "" ]; then
params="${params}, "
fi
params="${params}'$1'"
shift
done
cat pyGlue.py | sed -e "s/pyGlue/${name}/g" | \
sed -e "/_svc_name_ =/s?=.*?= '${name}'?" | \
sed -e "/_svc_display_name_ =/s?=.*?= '${display}'?" | \
sed -e "/_svc_description_ =/s?=.*?= '${desc}'?" | \
sed -e "/service_exe =/s?=.*?= '$exe'?" | \
sed -e "/service_script =/s?=.*?= '$script'?" | \
sed -e "/service_params =/s?=.*?= [${params}]?" | \
sed -e "/service_startDir =/s?=.*?= '${startDir}'?" > ${name}.py
cxfreeze ${name}.py --include-modules=win32timezone
Installation - copy the '.exe' the server and the script to the specified folder. Run the '.exe', as Administrator, with the 'install' option. Open Windows Services, as Adminstrator, and start you service. For upgrade, just copy the new version of the script and Stop/Start the service.
Now every server is different - different installations of Python, different folder structures. I maintain a folder for every server, with a copy of pyGlue.py and build.sh. And I create a 'serverBuild.sh' script for rebuilding all the service on that server.
# A script to build all the script based Services on this PC
sh build.sh AutoCode 'AutoCode Medical Documents' 'Autocode Medical Documents to SNOMED_CT and AIHW codes' C:/Python38/python.exe autocode.py C:/Users/russell/Documents/autocoding -S -T
The accepted answer using win32serviceutil works but is complicated and makes debugging and changes harder. It is far easier to use NSSM (the Non-Sucking Service Manager). You write and comfortably debug a normal python program and when it finally works you use NSSM to install it as a service in less than a minute:
From an elevated (admin) command prompt you run nssm.exe install NameOfYourService and you fill-in these options:
path: (the path to python.exe e.g. C:\Python27\Python.exe)
Arguments: (the path to your python script, e.g. c:\path\to\program.py)
By the way, if your program prints useful messages that you want to keep in a log file NSSM can also handle this and a lot more for you.
This doesn't answer the original question, but might help other people that want to automatically start a Python script at Windows startup:
Have a look at the Windows Task Scheduler instead, it is way easier if you just want to start a script after boot without all the service functionality of a Windows Service.
Create a new task, select "At startup" as trigger, "Start program" as action with "C:\Python39\python.exe" as the program (or wherever your python.exe is) and the full path to your script ("C:...\my_dir\xyz.py") as argument (you can use " if the path contains spaces).
You can also select the path of your script (without the .py file, e.g. "C:...\my_dir") for "start in" if you use relative paths in your script, e.g. for logging.
https://www.chrisumbel.com/article/windows_services_in_python
Follow up the PySvc.py
changing the dll folder
I know this is old but I was stuck on this forever. For me, this specific problem was solved by copying this file - pywintypes36.dll
From -> Python36\Lib\site-packages\pywin32_system32
To -> Python36\Lib\site-packages\win32
setx /M PATH "%PATH%;C:\Users\user\AppData\Local\Programs\Python\Python38-32;C:\Users\user\AppData\Local\Programs\Python\Python38-32\Scripts;C:\Users\user\AppData\Local\Programs\Python\Python38-32\Lib\site-packages\pywin32_system32;C:\Users\user\AppData\Local\Programs\Python\Python38-32\Lib\site-packages\win32
changing the path to python folder by
cd C:\Users\user\AppData\Local\Programs\Python\Python38-32
NET START PySvc
NET STOP PySvc