FAQ

Page Discussion Edit History

SimplePythonFCGI

Contents

[edit] Dispatching TurboGears Python via FCGI

This confirmed to run on Mac OS X 10.4.7 under Turbogears 0.9.9 and 1.1a (so, no reason not to run under the 1.0b release).

Information was drawn from the Turbogears trac wiki (http://trac.turbogears.org/turbogears/wiki/NginxIntegration) which show how to use Nginx to proxy to TG, and the nearby [NginxFcgiExample] page, the latter detailing the PHP/FCGI process.

This is for Nginx/FCGI/Turbogears

[edit] Conventions

Substitute thoughout with the values relevant to your own set-up:

${HOST} = localhost - (or whatever you choose)

${PORT} = 8080 - (or whatever you choose)

${NGINX} = /usr/local/nginx - location of nginx installation

${PROJECTBASE} /opt/projects/wiki20 - location of Turbogears project

${PROJECTNAME} wiki20

[edit] Getting the required files

Two files are required to be created: ${NGINX}/scripts/fcgi.py and ${NGINX}/scripts/${PROJECTNAME}.fcgi.

To create ${NGINX}/scripts/fcgi.py ...

mkdir ${NGINX}/scripts
curl -o ${NGINX}/scripts/fcgi.py http://www.saddi.com/software/py-lib/py-lib/fcgi.py

To create ${NGINX}/scripts/${PROJECTNAME}.fcgi ...

Copy and paste the following to ${NGINX}/scripts/${PROJECTNAME}.fcgi. Edit the file, navigate to the "USER EDIT SECTION" and replace each instance of ${PROJECTBASE} and ${PROJECTNAME} with the corresponding values for your project.

#!/usr/bin/python
#
# File name: project.fcgi
#
# This module provides the glue for running TurboGears applications behind
# FastCGI-enabled web servers. The code in this module depends on the fastcgi
# module downloadable from here:
#
# http://www.saddi.com/software/py-lib/py-lib/fcgi.py
#
# NOTE: The fcgi.py file needs to be placed in a location that is on the
# system path, such as the same the directory as the tg_fastcgi.py file
# or in the base directory of the TG app code.
#
# To configure this module, please edit the three variables in the "USER EDIT
# SECTION" before starting the TG application. Also remember to edit the
# top of this file with the correct Python installation information.
 
import cherrypy
import sys
import os
from os.path import *
import pkg_resources
import turbogears
 
pkg_resources.require("TurboGears")
 
# -- START USER EDIT SECTION
# -- Users must edit this section --
code_dir = '${PROJECTBASE}' # (Required) The base directory of the TG app code.
root_class_name = '${PROJECTNAME}.controllers.Root' # (Required) The fully qualified Root class name.
project_module_name = '${PROJECTNAME}.config' # (Required) The config module name.
log_dir = '' # (Optional) The log directory. Default = code_dir.
# -- END USER EDIT SECTION
 
class VirtualPathFilter(object):
    def on_start_resource(self):
        if not cherrypy.config.get('virtual_path_filter.on', False):
     return
 prefix = cherrypy.config.get('virtual_path_filter.prefix', '')
 if not prefix:
     return
 
 path = cherrypy.request.object_path
 if path == prefix:
     path = '/'
 elif path.startswith(prefix):
     path = path[len(prefix):]
 else:
     raise cherrypy.NotFound(path)
 cherrypy.request.object_path = path
 
 
def tg_init():
    """ Checks for the required data and initializes the application. """
 
    global code_dir
    global root_class_name
    global log_dir
    global project_module_name
    last_mark = 0
 
    # Input checks
    if not code_dir or not isdir(code_dir):
        raise ValueError("""The code directory setting is missing.
                            The fastcgi code will be unable to find
                            the TG code without this setting.""")
 
    if not root_class_name:
        raise ValueError("""The fully qualified root class name must
                            be provided.""")
 
    last_mark = root_class_name.rfind('.')
 
    if last_mark < 1 or last_mark + 1 == len(root_class_name):
        raise ValueError("""The user-defined class name is invalid.
                            Please make sure to include a fully
                            qualified class name for the root_class
                            value (e.g. wiki20.controllers.Root).""")
 
    sys.path.append(code_dir)
 
    # Change the directory so the TG log file will not be written to the
    # web app root.
    if log_dir and isdir(log_dir):
        os.chdir(log_dir)
    else:
        os.chdir(code_dir)
        log_dir = code_dir
 
    sys.stdout = open(join(log_dir, 'stdout.log'),'a')
    sys.stderr = open(join(log_dir, 'stderr.log'),'a')
 
    if exists(join(code_dir, "setup.py")):
        turbogears.update_config(configfile=join(code_dir, "dev.cfg"),modulename=project_module_name)
    else:
        turbogears.update_config(configfile=join(code_dir, "prod.cfg"),modulename=project_module_name)
 
    # Set environment to production to disable auto-reload and
    # add virutal path information.
    cherrypy.config.update({
        'global': {'server.environment': 'production'},
     '/' : { 'virtual_path_filter.on' : True,
     'virtual_path_filter.prefix' : '/bel.fcgi' }
  })
 
    # Parse out the root class information for Cherrypy Root class.
    package_name = root_class_name[:last_mark]
    class_name = root_class_name[last_mark+1:]
    _temp = __import__(package_name, globals(), locals(), [class_name], -1)
    Root = getattr(_temp, class_name)
    Root._cp_filters = [VirtualPathFilter()]
    cherrypy.root = Root()
 
# Main section -
# Initialize the application, then start the server.
tg_init()
 
from fcgi import WSGIServer
cherrypy.server.start(initOnly=True, serverClass=None)
 
from cherrypy._cpwsgi import wsgiApp
WSGIServer(application=wsgiApp).run()

[edit] Adjust the TurboGears configuration

Edit the ${PROJECTBASE}/dev.cfg or ${PROJECTBASE}/prod.cfg file (whichever you are using), uncomment the server.socket_port assignment and change ${PORT} to a value of your choice (make sure nothing else is running on that port, Tomcat defaults to 8080, as does Jetty. Save yourself some time and check first with a telnet localhost 8080, you should see Connection refused).

The relevant lines in prod/dev.cfg are:

server.socket_port=${PORT}

[edit] Spawning a FastCGI TurboGears process

The lighttpd "spawn-fcgi" script is useful: download, compile and install lighttpd. Then (replacing ${HOST} and ${PORT} values appropriately), execute the following:

/usr/local/bin/spawn-fcgi -a ${HOST} -p ${PORT} -u nobody -f ${NGINX}/scripts/${PROJECTNAME}.fcgi

[edit] Nginx configuration

Save the following into ${NGINX}/conf/fastcgi_params

#fastcgi.conf
fastcgi_param GATEWAY_INTERFACE CGI/1.1;
fastcgi_param SERVER_SOFTWARE nginx;
 
fastcgi_param QUERY_STRING $query_string;
fastcgi_param REQUEST_METHOD $request_method;
fastcgi_param CONTENT_TYPE $content_type;
fastcgi_param CONTENT_LENGTH $content_length;
 
fastcgi_param SCRIPT_NAME $fastcgi_script_name;
fastcgi_param REQUEST_URI $request_uri;
fastcgi_param DOCUMENT_URI $document_uri;
fastcgi_param DOCUMENT_ROOT $document_root;
fastcgi_param SERVER_PROTOCOL $server_protocol;
 
fastcgi_param REMOTE_ADDR $remote_addr;
fastcgi_param REMOTE_PORT $remote_port;
fastcgi_param SERVER_ADDR $server_addr;
fastcgi_param SERVER_PORT $server_port;
fastcgi_param SERVER_NAME $server_name;

Add the following to the server section of the ${NGINX}/conf/nginx.conf configuration file, changing ${HOST} and ${PORT} as appropriate:

  # static files
  location ~ ^/(images|javascript|js|css|flash|media|static)/ {
    root ${PROJECTBASE}/${PROJECTNAME}/static;
  }
 
  location = /favicon.ico {
    root ${PROJECTBASE}/${PROJECTNAME}/static/images;
  }
 
  # pass all requests to FastCGI TG server listening on ${HOST}:${PORT}
  #
  location / {
    fastcgi_pass ${HOST}:${PORT};
    fastcgi_index index;
    fastcgi_param SCRIPT_FILENAME /scripts$fastcgi_script_name;
    include conf/fastcgi_params;
 }

[edit] Starting nginx

Start nginx with ${NGINX}/sbin/nginx. Point your browser to http://${HOST}:${PORT}/, your Turboears project should be serving via FastCGI. If so ... congratulations.

[edit] Performance test software

Basic but usefully free http://www.hpl.hp.com/research/linux/httperf/

[edit] Note

[I left the IP address as 0.0.0.0 because it worked for me, whereas 127.0.0.1 did not. If you're experiencing difficulties connecting to 0.0.0.0:8080, these are both alternative options; localhost:8080, 127.0.0.1:8080.]

Good luck.