This post attempts to explain how to fork child processes in Python, or at least how to use forking on an existing Python script. For some strange reason I’ve had to explain this a few times recently, so I decided an easy to reference blog post would probably make life a little easier.
So you have a basic Python script that you’d like to run as a background process, for whatever reason; perhaps it’s a network service that waits for incoming connections, or you want to watch a log file for something. As an example, let’s say it looks something like this:
import time import sys def do_something(): # Do something for 10 seconds, then exit nicely. time.sleep(10) print "Done." sys.exit(0) if __name__ == "__main__": do_something()
So one method of running it in the background is to fork it. The term fork means that we want to launch a child process from this process (the parent process) and then disown that child process and exit cleanly. The child process will then continue running in the background until it completes whatever it has to do, or until it is shutdown or killed, in the case of a long running process.
In order to accomplish this we need something that looks like the following:
import time import sys import os def do_something(): # Do something for 10 seconds, then exist cleanly. time.sleep(10) print "kbye" sys.exit(0) if __name__ == "__main__": print ("Hello and welcome to a Python forking example. I'll now fork a " "backgrounded child process and then exit, leaving it to run all by " "itself. Watch for the 'Done' in 10 seconds...") try: pid = os.fork() if pid > 0: # Exit parent process sys.exit(0) except OSError, e: print >> sys.stderr, "fork failed: %d (%s)" % (e.errno, e.strerror) sys.exit(1) # Configure the child processes environment os.chdir("/") os.setsid() os.umask(0) # Execute something in the background do_something()
So what is actually going on here ?
Nothing changes in the actual code apart from giving it the ability to fork. This makes it easy to give existing (or 3rd party) code the ability to fork.
First off, we execute a system fork() call (which is an operating system call), this returns the process ID of the child process. We then have some basic sanity checking and make sure it’s a real PID and the forking didn’t raise any exceptions.
After that, we change the child processes environment so that it is no longer a child of our main parent process and so that it can live on its own after we exit.
Finally, we execute the original code, now inside a detached child process. This can run for as long as it needs, and could potentially log to a log file if you care about any stdout output.
There are also some better alternatives, something like this for example:
import os os.spawnl(os.P_DETACH, 'some_log_running_command')
os.P_DETACH is win32 specific. For better portability you could use