Skip to main content

Unicorn Herder: manage daemonized (g)unicorns

Project description

Unicorn and Gunicorn are awesome tools for people writing web services in Ruby and Python. One of the more nifty features of both programs is their ability to reload application code on-the-fly, by spawning a new master process (or “arbiter”, in Gunicorn’s language) in response to an operating system signal (SIGUSR2). Unfortunately, this reloading process is incompatible with process-tracking supervisors such as Upstart, because the old master process dies as part of the reload.

Unicorn Herder is a utility designed to assist in the use of Upstart and similar supervisors with Unicorn. It does this by polling the pidfile written by the Unicorn master process, and automating the sequence of signals that must be sent to the master to do a “hot-reload”. If Unicorn quits, so will the Unicorn Herder, meaning that if you supervise the herder (which does not daemonize), you are effectively supervising the Unicorn process.


Unicorn Herder is available from the Python Package Index, and can be installed with pip:

$ pip install unicornherder


With gunicorn:

$ unicornherder -- -w 4 myapp:app

With unicorn (using Bundler):

$ bundle exec unicornherder -u unicorn


Unicorn Herder forwards the following signals to the unicorn master process:


Notably, Unicorn Herder does not forward SIGWINCH, because it is not intended to be daemonized.

Unicorn Herder also intercepts SIGHUP, because this is the signal sent by Upstart when you call initctl reload, and uses it to trigger a hot-reload of its Unicorn instance. This process will take two minutes by default, in order to give the new workers time to start up.

NB: There will be a period during hot-reload when requests are served by both old and new workers. This might have serious implications if you are running data migrations between deploying versions of your application. Please bear this in mind when deciding if you should use Unicorn Herder’s hot-reloading feature.

This diagram shows the interactions between unicornherder and (g)unicorn masters at boot and around hot-reloading with HUP (ignoring workers):


Upstart config

An example upstart config (compatible with Upstart v1.4 and later) for use with Unicorn Herder is given below:

description "Unicorn Herder"

start on runlevel [2345]
stop on runlevel [!2345]

respawn limit 5 20

env PORT=4567

setuid www
setgid www

chdir /var/apps/myapp

exec bundle exec unicornherder -u unicorn -- --port $PORT

# Or, for a gunicorn installation with a virtualenv
# at /var/venv/myapp...

#  . /var/venv/myapp/bin/activate
#  exec unicornherder -- -w 4 -b "$PORT" myapp:app
#end script


You can discuss this tool with our open source mailing list:


Unicorn Herder is released under the MIT license, a copy of which can be found in LICENSE.


To upload a new version of unicornherder to PyPI, you can follow these instructions:

$ python sdist $ pip install twine $ twine upload dist/*

The credentials for the GDS PyPI account are available in govuk-secrets.

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

unicornherder-0.2.1.tar.gz (11.1 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