Skip to main content

A Python library for creating super fancy Unix daemons

Project description

daemonocle is a library for creating your own Unix-style daemons written in Python. It solves many problems that other daemon libraries have and provides some really useful features you don’t often see in other daemons.


To install via pip:

pip install daemonocle

Or download the source code and install manually:

git clone
cd daemonocle/
python install

Basic Usage

Here’s a really really basic example:

import sys
import time

import daemonocle

# This is your daemon. It sleeps, and then sleeps again.
def main():
    while True:

if __name__ == '__main__':
    daemon = daemonocle.Daemon(

And here’s the same example with logging and a Shutdown Callback:

import logging
import sys
import time

import daemonocle

def cb_shutdown(message, code):'Daemon is stopping')

def main():
        level=logging.DEBUG, format='%(asctime)s [%(levelname)s] %(message)s',
    )'Daemon is starting')
    while True:
        logging.debug('Still running')

if __name__ == '__main__':
    daemon = daemonocle.Daemon(

And here’s what it looks like when you run it:

user@host:~$ python start
Starting ... OK
user@host:~$ python status -- pid: 1234, status: running, uptime: 1m, %cpu: 0.0, %mem: 0.0
user@host:~$ python stop
Stopping ... OK
user@host:~$ cat /var/log/daemonocle_example.log
2014-05-04 12:39:21,090 [INFO] Daemon is starting
2014-05-04 12:39:21,091 [DEBUG] Still running
2014-05-04 12:39:31,091 [DEBUG] Still running
2014-05-04 12:39:41,091 [DEBUG] Still running
2014-05-04 12:39:51,093 [DEBUG] Still running
2014-05-04 12:40:01,094 [DEBUG] Still running
2014-05-04 12:40:07,113 [INFO] Daemon is stopping
2014-05-04 12:40:07,114 [DEBUG] Terminated by SIGTERM (15)

For more details, see the Detailed Usage section below.


If you think about it, a lot of Unix daemons don’t really know what the hell they’re doing. Have you ever found yourself in a situation that looked something like this?

user@host:~$ sudo example start
starting example ... ok
user@host:~$ ps aux | grep example
user      1234  0.0  0.0   1234  1234 pts/1    S+   12:34   0:00 grep example
user@host:~$ sudo example start
starting example ... ok
user@host:~$ echo $?
user@host:~$ tail -f /var/log/example.log

Or something like this?

user@host:~$ sudo example stop
stopping example ... ok
user@host:~$ ps aux | grep example
user       123  0.0  0.0   1234  1234 ?        Ss   00:00   0:00 /usr/local/bin/example
user      1234  0.0  0.0   1234  1234 pts/1    S+   12:34   0:00 grep example
user@host:~$ sudo example stop
stopping example ... ok
user@host:~$ ps aux | grep example
user       123  0.0  0.0   1234  1234 ?        Ss   00:00   0:00 /usr/local/bin/example
user      1240  0.0  0.0   1234  1234 pts/1    S+   12:34   0:00 grep example
user@host:~$ sudo kill -9 123

Or something like this?

user@host:~$ sudo example status
Usage: example {start|stop|restart}
user@host:~$ ps aux | grep example

These are just a few examples of unnecessarily common problems. It doesn’t have to be this way.

Note: You might be thinking, “Why not just write a smarter start/stop shell script wrapper for your daemon that checks whether or not it actually started, actually stopped, etc.?” Seriously? It doesn’t have to be this way. I believe daemons should be more self-aware. They should handle their own problems most of the time, and your start/stop script should only be a very thin wrapper around your daemon or simply a symlink to your daemon.

The Problem

If you’ve ever dug deep into the nitty-gritty details of how daemonization works, you’re probably familiar with the standard “double fork” paradigm first introduced by W. Richard Stevens in the book Advanced Programming in the UNIX Environment. One of the problems with the standard way to implement this is that if the final child dies immediately when it gets around to doing real work, the original parent process (the one that actually had control of your terminal) is long gone. So all you know is that the process got forked, but you have no idea if it actually kept running for more than a fraction of a second. And let’s face it, one of the most likely times for a daemon to die is immediately after it starts (due to bad configuration, permissions, etc.).

The next problem mentioned in the section above is when you try to stop a daemon, it doesn’t actually stop, and you have no idea that it didn’t actually stop. This happens when a process doesn’t respond properly to a SIGTERM signal. It happens more often than it should. The problem is not necessarily the fact that it didn’t stop. It’s the fact that you didn’t know that it didn’t stop. The start/stop script knows that it successfully sent the signal and so it assumes success. This also becomes a problem when your restart command blindly calls stop and then start, because it will try to start a new instance of the daemon before the previous one has exited.

These are the biggest problems most daemons have in my opinion. daemonocle solves these problems and provides many other “fancy” features.

The Solution

The problem with the daemon immediately dying on startup and you not knowing about it is solved by having the first child (the immediate parent of the final child) sleep for one second and then call os.waitpid(pid, os.WNOHANG) to see if the process is still running. This is what daemonocle does. So if you’re daemon dies within one second of starting, you’ll know about it.

This problem with the daemon not stopping and you not knowing about it is solved by simply waiting for the process to finish (with a timeout). This is what daemonocle does. (Note: When a timeout occurs, it doesn’t try to send a SIGKILL. This is not always what you’d want and often not a good idea.)

Other Useful Features

Below are some other useful features that daemononcle provides that you might not find elsewhere.

The status Action

There is a status action that not only displays whether or not the daemon is running and its PID, but also the uptime of the daemon and the % CPU and % memory usage of all the processes in the same process group as the daemon (which are probably its children). So if you have a daemon that launches mulitple worker processes, the status action will show the % CPU and % memory usage of all the workers combined.

It might look something like this:

user@host:~$ python status -- pid: 1234, status: running, uptime: 12d 3h 4m, %cpu: 12.4, %mem: 4.5

Slightly Smarter restart Action

Have you ever tried to restart a daemon only to realize that it’s not actually running? Let me guess: it just gave you an error and didn’t start the daemon. A lot of the time this is not a problem, but if you’re trying to restart the daemon in an automated way, it’s more annoying to have to check if it’s running and do either a start or restart accordingly. With daemonocle, if you try to restart a daemon that’s not running, it will give you a warning saying that it wasn’t running and then start the daemon. This is often what people expect.


Daemons that use daemonocle have the ability to reload themselves by simply calling daemon.reload() where daemon is your daemonocle.Daemon instance. The execution of the current daemon halts wherever daemon.reload() was called, and a new daemon is started up to replace the current one. From your code’s perspective, it’s pretty much the same as a doing a restart except that it’s initiated from within the daemon itself and there’s no signal handling involved. Here’s a basic example of a daemon that watches a config file and reloads itself when the config file changes:

import os
import sys
import time

import daemonocle

class FileWatcher(object):

    def __init__(self, filename, daemon):
        self._filename = filename
        self._daemon = daemon
        self._file_mtime = os.stat(self._filename).st_mtime

    def file_has_changed(self):
        current_mtime = os.stat(self._filename).st_mtime
        if current_mtime != self._file_mtime:
            self._file_mtime = current_mtime
            return True
        return False

    def watch(self):
        while True:
            if self.file_has_changed():

if __name__ == '__main__':
    daemon = daemonocle.Daemon(pidfile='/var/run/')
    fw = FileWatcher(filename='/etc/daemonocle_example.conf', daemon=daemon)
    daemon.worker =

Shutdown Callback

You may have noticed from the Basic Usage section above that a shutdown_callback was defined. This function gets called whenever the daemon is shutting down in a catchable way, which should be most of the time except for a SIGKILL or if your server crashes unexpectedly or loses power or something like that. This function can be used for doing any sort of cleanup that your daemon needs to do. Also, if you want to log (to the logger of your choice) the reason for the shutdown and the intended exit code, you can use the message and code arguments that will be passed to your callback (your callback must take these two arguments).

Non-Detached Mode

This is not particularly interesting per se, but it’s worth noting that in non-detached mode, your daemon will do everything else you’ve configured it to do (i.e. setuid, setgid, chroot, etc.) except actually detaching from your terminal. So while you’re testing, you can get an extremely accurate view of how your daemon will behave in the wild. It’s also worth noting that self-reloading works in non-detached mode, which was a little tricky to figure out initially.

File Descriptor Handling

One of the things that daemons typically do is close all open file descriptors and establish new ones for STDIN, STDOUT, STDERR that just point to /dev/null. This is fine most of the time, but if your worker is an instance method of a class that opens files in its __init__() method, then you’ll run into problems if you’re not careful. This is also a problem if you’re importing a module that leaves open files behind. For example, importing the random standard library module in Python 3 results in an open file descriptor for /dev/urandom.

Since this “feature” of daemons often causes more problems than it solves, and the problems it causes sometimes have strange side-effects that make it very difficult to troubleshoot, this feature is optional and disabled by default in daemonocle via the close_open_files option.

Detailed Usage

The daemonocle.Daemon class is the main class for creating a daemon using daemonocle. Here’s the constructor signature for the class:

class daemonocle.Daemon(
    worker=None, shutdown_callback=None, prog=None, pidfile=None, detach=True,
    uid=None, gid=None, workdir='/', chrootdir=None, umask=022, stop_timeout=10,

And here are descriptions of all the arguments:


The function that does all the work for your daemon.


This will get called anytime the daemon is shutting down. It should take a message and a code argument. The message is a human readable message that explains why the daemon is shutting down. It might useful to log this message. The code is the exit code with which it intends to exit. See Shutdown Callback for more details.


The name of your program to use in output messages. Default: os.path.basename(sys.argv[0])


The path to a PID file to use. It’s not required to use a PID file, but if you don’t, you won’t be able to use all the features you might expect. Make sure the user your daemon is running as has permission to write to the directory this file is in.


Whether or not to detach from the terminal and go into the background. See Non-Detached Mode for more details. Default: True


The user ID to switch to when the daemon starts. The default is not to switch users.


The group ID to switch to when the daemon starts. The default is not to switch groups.


The path to a directory to change to when the daemon starts. Note that a file system cannot be unmounted if a process has its working directory on that file system. So if you change the default, be careful about what you change it to. Default: "/"


The path to a directory to set as the effective root directory when the daemon starts. The default is not to do anything.


The file creation mask (“umask”) for the process. Default: 022


Number of seconds to wait for the daemon to stop before throwing an error. Default: 10


Whether or not to close all open files when the daemon detaches. Default: False


The default actions are start, stop, restart, and status. You can get a list of available actions using the daemonocle.Daemon.list_actions() method. The recommended way to call an action is using the daemonocle.Daemon.do_action(action) method. The string name of an action is the same as the method name except with dashes in place of underscores.

If you want to create your own actions, simply subclass daemonocle.Daemon and add the @daemonocle.expose_action decorator to your action method, and that’s it.

Here’s an example:

import daemonocle

class MyDaemon(daemonocle.Daemon):

    def full_status(self):
        """Get more detailed status of the daemon."""

Then, if you did the basic daemon.do_action(sys.argv[1]) like in all the examples above, you can call your action with a command like python full-status.

Integration with mitsuhiko’s click

daemonocle also provides an integration with click, the “composable command line utility”. The integration is in the form of a custom command class daemonocle.cli.DaemonCLI that you can use in conjunction with the @click.command() decorator to automatically generate a command line interface with subcommands for all your actions. It also automatically daemonizes the decorated function. The decorated function becomes the worker, and the actions are automatically mapped from click to daemonocle.

Here’s an example:

import time

import click
from daemonocle.cli import DaemonCLI

@click.command(cls=DaemonCLI, daemon_params={'pidfile': '/var/run/'})
def main():
    """This is my awesome daemon. It pretends to do work in the background."""
    while True:

if __name__ == '__main__':

Running this example would look something like this:

user@host:~$ python --help
Usage: [<options>] <command> [<args>]...

  This is my awesome daemon. It pretends to do work in the background.

  --help  Show this message and exit.

  start    Start the daemon.
  stop     Stop the daemon.
  restart  Stop then start the daemon.
  status   Get the status of the daemon.
user@host:~$ python start --help
Usage: start [<options>]

  Start the daemon.

  --debug  Do NOT detach and run in the background.
  --help   Show this message and exit.

The daemonocle.cli.DaemonCLI class also accepts a daemon_class argument that can be a subclass of daemonocle.Daemon. It will use your custom class, automatically create subcommands for any custom actions you’ve defined, and use the docstrings of the action methods as the help text just like click usually does.

This integration is entirely optional. daemonocle doesn’t enforce any sort of argument parsing. You can use argparse, optparse, or just plain sys.argv if you want.

Bugs, Requests, Questions, etc.

Please create an issue on GitHub.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

xdaemonocle-0.8.1.tar.gz (18.0 kB view hashes)

Uploaded source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page