An IIS-Python bridge based on WSGI and FastCGI.
wfastcgi.py provides a bridge between IIS and Python using WSGI and FastCGI, similar to what mod_python provides for Apache HTTP Server.
It can be used with any Python web application or framework that supports WSGI, and provides an efficient way to handle requests and process pools through IIS.
To install via the Python Package Index (PyPI), type:
pip install wfastcgi
Installing IIS and FastCGI
See the IIS Installation page for information about installing IIS on your version of Windows.
The Application Development/CGI package is required for use with
Once wfastcgi and IIS are installed, run wfastcgi-enable as an administrator to enable wfastcgi in the IIS configuration. This will configure a CGI application that can then be specified as a route handler.
To disable wfastcgi before uninstalling, run wfastcgi-disable.
wfastcgi-disable pip uninstall wfastcgi
Note: uninstalling wfastcgi does not automatically unregister the CGI application.
If the first argument passed to wfastcgi-enable or wfastcgi-disable is a valid file, the entire command line is used to register or unregister the CGI handler.
For example, the following command will enable wfastcgi with IIS Express and a specific host configuration:
wfastcgi-enable "C:\Program Files (x86)\IIS Express\appcmd.exe" /apphostconfig:C:\Path\To\applicationhost.config
You can disable wfastcgi in the same configuration file using wfastcgi-disable with the same options:
wfastcgi-disable "C:\Program Files (x86)\IIS Express\appcmd.exe" /apphostconfig:C:\Path\To\applicationhost.config
Routing requests to your Python application requires some site-local configuration. In your site’s web.config file, you will need to add a handler and some app settings:
<configuration> <system.webServer> <handlers> <add name="Python FastCGI" path="*" verb="*" modules="FastCgiModule" scriptProcessor="C:\Python36\python.exe|C:\Python36\Lib\site-packages\wfastcgi.py" resourceType="Unspecified" requireAccess="Script" /> </handlers> </system.webServer> <appSettings> <!-- Required settings --> <add key="WSGI_HANDLER" value="my_app.wsgi_app()" /> <add key="PYTHONPATH" value="C:\MyApp" /> <!-- Optional settings --> <add key="WSGI_LOG" value="C:\Logs\my_app.log" /> <add key="WSGI_RESTART_FILE_REGEX" value=".*((\.py)|(\.config))$" /> <add key="APPINSIGHTS_INSTRUMENTATIONKEY" value="__instrumentation_key__" /> <add key="DJANGO_SETTINGS_MODULE" value="my_app.settings" /> <add key="WSGI_PTVSD_SECRET" value="__secret_code__" /> <add key="WSGI_PTVSD_ADDRESS" value="ipaddress:port" /> </appSettings> </configuration>
The value for scriptProcessor is displayed in the output of wfastcgi-enable and may vary from machine to machine. The values for path and verb may also be customized to further restrict the requests for which this handler will be used.
The name value may be used in nested web.config files to exclude this handler. For example, adding a web.config to your static/ subdirectory containing <remove name="Python FastCGI" /> will prevent IIS from serving static files through your Python app.
The provided app settings are translated into environment variables and can be accessed from your Python application using os.getenv. The following variables are used by wfastcgi.
This is a Python name that evaluates to the WSGI application object. It is a series of dotted names that are optionally called with no parameters. When resolving the handler, the following steps are used:
- As many names as possible are loaded using import. The last name is never imported.
- Once a module has been obtained, each remaining name is retrieved as an attribute. If () follows the name, it is called before getting the following name.
Errors while resolving the name are returned as a simple 500 error page. Depending on your IIS configuration, you may only receive this page when accessing the site from the same machine.
Python is already running when this setting is converted into an environment variable, so wfastcgi performs extra processing to expand environment variables in its value (including those added from app settings) and to expand sys.path.
If you are running an implementation of Python that uses a variable named something other than PYTHONPATH, you should still specify this value as PYTHONPATH.
This is a full path to a writable file where logging information is written. This logging is not highly efficient, and it is recommended that this setting only be specified for debugging purposes.
The regular expression used to identify when changed files belong to your website. If a file belonging to your site changes, all active CGI processes will be terminated so that the new files can be loaded.
By default, all *.py and *.config files are included. Specify an empty string to disable auto-restart.
Providing an instrumentation key with this value will enable request tracing with Application Insights for your entire site. If you have not installed the applicationinsights package, a warning is written to WSGI_LOG (if enabled) but the site will operate normally.
Application Insights is a low-overhead monitoring system for tracking your application’s health and performance. When enabled, all errors in your site will be reported through Application Insights.
A commonly used registry key when deploying sites built using Django. Typically Django sites will set WSGI_HANDLER to django.core.handlers.wsgi.WSGIHandler() and load app-specific settings through the module specified by this value.
Sites using frameworks other than Django do not need to specify this value.
Providing an arbitrary string here and including the ptvsd module in your environment will automatically enable remote debugging of your web site. The string in this application setting should be treated as a password, and needs to be provided when attaching to the running site.
When WSGI_PTVSD_SECRET is specified, this value may also be specified to override the default listening address for remote debugging. By default, your site will listen on localhost:5678, but in many cases you may need to change this to 0.0.0.0:some-port in order to attach remotely.
Remember that you will also need to forward the port through any firewalls you might have configured.