version 0.8.0alpah is updated for django 1.5 and now includes autoreloading of files when then change and integration with django's static files application.
django-wsgiserver is a django app for serving Django sites via CherryPy's excellent, production ready, pure-python web server without needing to install all of Cherrypy. Note that Cherrypy names its server "wsgiserver" but it is actually a full-blown, multi-threaded http/https web server which has been used on production sites alone, or more often proxied behind a something like Apache or nginx.
The wsgiserver component has been used for years in production. Peter Baumgartner noted that it solved problems for him on memory on a memory-limited VPS hosted site . Performance-wise it does well: it can serve up thousands of requests per second [Piel2010].
I have used django-wsgiserver for small production sites, though daemonized modwsgi and uwsgi have served me well. I use it all the time though during development. It's my "pocket-sized" server. completely written in python and it gives me an instant approximation of the final production environment I use. In some ways, it's much better than the development server which is built into django: It gives a quick way of running an SSL for sites that require https, and it has no problem serving pages that do multiple ajax calls that would cause the built-in runserver to hang.
To use django-wsgiserver, I add it to my installed apps in my project settings file, then do:
$ manage.py runwsgiserver
and reload my browser page and the problem is fixed. It's also useful to see if some weird effect is being caused by runserver's process of loading the settings twice.
This project is a slight modification of code form Peter Baumgartner code (see Peter's blog post) Peter and others did the work of creating a management command. I've added a few small improvements from my point of view: It doesn't require installing cherrypy separately. It uses the same port as the development server (8000) so I don't need to re-enter my testing url in my browser, and it works by default with OS's like Mac OS 10.6 and 10.8 and Ubuntu 10.04 which prefer binding localhost to an ip6 address.
- multi-threaded production ready webserver for low to medium traffic sites
- pure python
- setuptools/pip installable (so install right in virtualenv)
- small memory footprint
- by default replaces mimics the django built-in server, serves admin media by default for easy testing and deployment. Will serve from STATICFILES_ROOT if it's defined.
- autoreloads code like the development server by default
- supports https/ssl to make testing under https easy
- It would be nice allow to run without installing into INSTALLED_APPS so don't need to touch code cf. gunicorn_django
- It would be nice to print out the requests in development mode (like runserver dose) using logging. Maybe do this with wsgi middleware?
- django version 1.5. It might work with 1.3 or 1.4. It will not work fully with 1.2 or below. (see 0.6 series for that)
To get started using the server, you need nothing outside of itself and the project code that you would like to serve up. However, for ssl support, you may need PyOpenSSL--though the new cherrypy server includes support for using the python built-in ssl module depending on which version of python you are using.
django-wsgiserver is BSD licensed based on lincolnloop's django-cpservers original code.
To install, django-wsgiserver follows the usual pattern for a django python application. You have several options
- pip install django-wsgiserver OR
- easy_install django-wsgiserver
- If you would like to use the latest possible version, you can use pip and mercurial checkout from bitbucket
pip install -e hg+https://bitbucket.org/cleemesser/django-wsgiserver#egg=django-wsgiserver
- Alternatively you can download the code and install it so that django_wsgiserver is on your PYTHONPATH
After you used one of the methods above, you need to add django_wsgiserver to your INSTALLED_APPS in your django project's settings file
To see how to run the server as a management command, run:
$ python manage.py runwsgiserver help
from within your project directory. You'll see something like what's below:
CPWSGI_HELP = r""" Run this project in CherryPy's production quality http webserver. Note that it's called wsgiserver but it is actually a complete http server. runwsgiserver [options] [cpwsgi settings] [stop] Optional CherryPy server settings: (setting=value) host=HOSTNAME hostname to listen on Defaults to 127.0.0.1, (set to 0.0.0.0 to bind all ip4 interfaces or :: for all ip6 interfaces) port=PORTNUM port to listen on Defaults to 8000 server_name=STRING CherryPy's SERVER_NAME environ entry Defaults to localhost daemonize=BOOL whether to detach from terminal Defaults to False pidfile=FILE write the spawned process-id to this file workdir=DIRECTORY change to this directory when daemonizing threads=NUMBER Number of threads for server to use ssl_certificate=FILE SSL certificate file ssl_private_key=FILE SSL private key file server_user=STRING user to run daemonized process Defaults to www-data server_group=STRING group to daemonized process Defaults to www-data staticserve=True|False|collectstatic] If True, serve the static files automatically using django.contrib.staticfiles like the builting django server. If staticserve=collectstatic, instead serve static files from a single directory at STATIC_ROOT. You need to run "manage.py collectstatic" first. Defaults to True. adminserve=True|False Deprecated. Has no effect. The admin is served if staticserve is active. Examples: Run a "standard" CherryPy wsgi server--good for local development $ manage.py runwsgiserver Run a CherryPy server on port 80 $ manage.py runwsgiserver port=80 Run a CherryPy server as a daemon and write the spawned PID in a file, don't serve staticfiles or autoreload $ manage.py runwsgiserver daemonize=true pidfile=/var/run/django-cpwsgi.pid autoreload=False staticserve=False Run a CherryPy server using ssl with test certificates located in /tmp $ manage.py runwsgiserver ssl_certificate=/tmp/testserver.crt ssl_private_key=/tmp/testserver.key Run the wsgi server but serve all the static files from a single collected file tree $ manage.py collectstatic # collects all the static files to STATIC_ROOT $ manage.py runwsgiserver staticserve=collectstatic
If you want to use an installed version of Cherrypy--perhaps because you now have a more recent version, you only need to change one line of code in (around line 177) of django_wsgiserver/management/commands/run_wsgiserver.py:
from django_wsgiserver.wsgiserver import CherryPyWSGIServer as Server #from cherrypy.wsgiserver import CherryPyWSGIServer as Server
Just comment out the import from django_wsgiserver.wsgiserver and uncomment the import from cherrypy.wsgiserver to make the switch. For SSL use, you need to search and replace "django_wsigserver." back to "cherrypy."
- I should probably just add a switch to allow use of the native cherrypy install
- Consider comparisons to other server backends: tornado, uwsgi, gunicorn
- further security tests
- add shell command that will serve a django project in a default way without needing to alter settings.py to add to INSTALLED_APPS
- ? add ability to printout requests like django built in runserver (cf. django-odeon)
- 0.8.0 target django 1.5+, python 2.6, 2.7, and 3.2+
- add autoreload as default
- serve static files correctly
- 0.7.0 target django 1.3 -- not released --
added support for serving static files directory by default if STATICFILES_ROOT is defined. Added more tests: ssl, staticfiles and started to document them. Experimental: STATICFILES_DIRS during development to avoid needing to collect all files. - Added django-odeon's patch to wsgiserver to make it robust to illegal header lines
0.6.9 typo fix
0.6.8 Changed name of bitbucket repo to django-wsgiserver to match it's pypi name
0.6.7 using open().read() in setup file broke setuptools/pip install because README.rst wasn't included. Created MANIFEST.in file and now include README.rst tests/, docs/
0.6.6 fix up cherrypy dependency in ssl that was accidently introduced in the move to cherrypy.wsgiserver 3.2 branch
0.6.5 added mediahandler wsgi application
9/6/10 0.6.4 added code to automatically serve the admin media like the development server does by default. Can turn off on command line for production.
9/6/10 0.6.3 - see if I can get the download to finally include all the packages--didn't have wsgiserver!
added test project in tests/ directory
got tired of typing run_cp_wsgiserver so did a rename so I could use runwsgiserver instead.
updated wsgiserver to svn r2680 (matches cherrypy version 3.2 beta+). This fixes some bugs and gives better python 2.6 support. This version of cherrypy will also support python 3.x for whenever django starts supporting it.
use port 8000 as with django devserver rather than Cherrypy's default 8088
adapted defaults host=127.0.0.1 in order to work with ip4 by default. This addresses an issue I first noticed on mac OS 10.6 and later on ubuntu 10.04 where ip6 is active by default. Can get around this by adjusting the host option. For binding all ip4 interfaces, set to 0.0.0.0. For all ip6 interfaces I believe you use '::' You can also bind a specific interface by specifying host=<specific ip address> See http://www.cherrypy.org/ticket/711
switched code to use run_cp_wsgiserver instead of runcpserver
Many thanks to Peter and lincolnloop for describing how to do this and writing the code.
Peter acknowledged idea and code snippets borrowed from Loic d'Anterroches, adapted to run as a management command
Note, there is also similar code on PyPi and at http://hg.piranha.org.ua/cpserver/ maintained by Alexander Solovyov
The latest version of the cherrypy wsgiserver can be retrieved with:
svn co http://svn.cherrypy.org/trunk/cherrypy/wsgiserver
Peter hosts his code at http://github.com/lincolnloop/django-cpserver
- Charl Botha (cpbotha) http://timescapers.com/ staticfiles work
- jamalex patch to prevent static files access outside of media root