The Apache Web Server module Mod-WSGI provides a high-performance environment in which to run TurboGears 2.1. It is part of the Standard Deployment Pattern which is recommended for new TurboGears users.
$ sudo aptitude install apache2 libapache2-mod-wsgi
Mod-WSGI will create a new type of executable script and add a number of Apache configuration directives which allow you to configure your Mod-WSGI environment. While you could, technically, use these scripts as simple CGI-like scripts, we recommend using a VirtualEnv based deployment pattern to allow you to install different versions of packages for different applications or sites.
The recommended pattern for using Mod-WSGI with a VirtualEnv is to create a “baseline” VirtualEnv which contains no packages at all. This “baseline” will be used to provide a “clean” environment on top of which your application’s environment will be layered. (This done with the Apache WSGIPythonHome directive).
$ sudo mkdir /usr/local/pythonenv $ cd /usr/local/pythonenv $ sudo virtualenv --no-site-packages BASELINE $ sudo chown -R www-data:www-data BASELINE
The “baseline” pattern works around limitations in the Mod-WSGI module which have been fixed in later versions of the module. However, the commonly-packaged versions of the module require these work-arounds to work reliably, so for now, we recommend this pattern for all new users.
Now you can create your application-specific VirtualEnv (normally in the same pythonenv directory).
$ sudo virtualenv --no-site-packages myapp $ sudo chown -R www-data:www-data myapp $ sudo -u www-data bash $ cd myapp $ source bin/activate (myapp)$ easy_install -i http://www.turbogears.org/2.1/downloads/current/index tg.devtools (myapp)$ mkdir python-eggs (myapp)$ exit
The python-eggs directory is used by Mod-WSGI to unpack “zipped” eggs which contain binary extensions (.so files). Python cannot load those packages unless they are on the file-system. Configuring this directory will be automatically done in the .wsgi file created by the deploy_modwsgi_deploy.
Your application’s lib/pythonX.X/site-packages directory will be added to the WSGI environment’s path via a WSGIPythonPath directive, and will be moved to the front of the PythonPath by the .wsgi script in order to make it the dominant source for packages.
You can use any VirtualEnv installation mechanism you like to set up the application’s VirtualEnv. See Alternate Installation Process for other options such as PIP. In particularly you may want to use a mechanism that allows you to explicitly control which packages are installed and keep local copies of them to prevent external dependencies.
You have a number of options for deploying your code. See Deploying Your Project Code.
If you are following this Standard Deployment Pattern, you should be sure to deploy the code of your project to /usr/local/turbogears/myapp.
If you are going to Deploy with a Source Code Checkout, make sure to install /usr/local/turbogears/myapp‘s dependencies (into /usr/local/pythonenv/myapp) and generate its meta-data by using the develop option to /usr/local/turbogears/myapp/setup.py.
While you can generate your Apache site-configuration files by hand, new users will generally find this a somewhat daunting task. The modwsgideploy project provides a small helper script which has parameterized helper scripts which can generate an initial Apache Mod-WSGI configuration file and .wsgi script.
$ cd ~ $ easy_install bzr $ bzr branch http://bazaar.launchpad.net/~mcfletch/modwsgideploy/parameterized/ $ cd parameterized/trunk $ source /usr/local/pythonenv/bin/activate (myapp)$ python setup.py develop (myapp)$ paster modwsgi_deploy --help (myapp)$ (myapp)$ cd /usr/local/turbogears/myapp (myapp)$ paster modwsgi_deploy --logging [other options you choose] (myapp)$ deactivate
When we have the branch integrated, replace with easy_install modwsgideploy
the script is heavily parameterized to allow you to configure your site as desired. If you want your site to be available as a sub-directory of your main site, you can specify a mount-point (the default is /projectname). If you want to set up VirtualHost support (where your server looks at the requested host-name to determine which site to display), you can specify the server-name on the command-line.
The script will create a directory (by default ./apache) which will contain the .wsgi script and an Apache configuration file. It will also (likely) log a number of warnings telling you how to create your Baseline VirtualEnv, your Application-Specific VirtualEnv, where to copy/checkout your project code, and where to put your production config file.
The files generated will look like this:
myapp |-- apache | |-- README.txt | |-- myapp | |-- myapp.wsgi | `-- test.wsgi
You should review and/or edit the generated files. See the References for documentation on the contents of these files.
The config files assume that your application is deployed in the deployment location (/usr/local/turbogears/ by default) in a directory named myapp with the config-file (production.ini by default) in that directory. The application’s directory will be added to the PYTHONPATH, as will the VirtualEnv’s directory.
When you are finished, you can continue on to Enable Your Apache Site.
If you have used print statements anywhere in your codebase, you can expect your Mod-WSGI applications to crash. Mod-WSGI will error out if there is any attempt to write to stdout (which is what print does by default). Use the logging module instead of print throughout your codebase.
In multiple process load balanced deployments (such as this one) it is very possible that a given request will pull resources from multiple processes.
You may want to make sure that the TG controllers are loaded up even before the first request comes in to handle this, so you should add:
import paste.fixture app = paste.fixture.TestApp(application) app.get("/")
to the end of the wsgi-script that starts your application.
This will fetch the index page of your app, thus assuring that it’s ready to handle all of your requests immediately. This avoids a problem where your controller page is not yet loaded so widgets aren’t initialized, but a request comes in for a widget resource the ToscaWidgets middleware doesn’t have registered yet.