lurkingideas

Middleware

First, you have to collect profiling data with cProfile. Plug it into your Django project with a middleware class, such as this:

from datetime import datetime
import cProfile


class InstrumentMiddleware(object):
    def process_request(self, request):
        if 'profile' in request.COOKIES:
            request.profiler = cProfile.Profile()
            request.profiler.enable()

    def process_response(self, request, response):
        if hasattr(request, 'profiler'):
            request.profiler.disable()
            stamp = (request.META['REMOTE_ADDR'], datetime.now())
            request.profiler.dump_stats('/tmp/%s-%s.pro' % stamp)
        return response

This will start collecting data in process_request and stop collecting data in process_response thus monitoring the whole request. This includes all other middleware or some middleware depending where you put it in your settings.py:

MIDDLEWARE_CLASSES = (
    ...
    'project.apps.myapp.middleware.InstrumentMiddleware',
    ...

Profile data

Next, collect the data for the page you want to profile. You will have to create a cookie in your browser called profile with whatever content. This is of course to prevent other pages from other users being profiled. You will find the profile data in the directory tmp on the server. Copy the file to your desktop.

Visualize

Install KCachegrind. I have not been able to find a substitute for this application so you will have to use that with all its KDE dependencies. Like this on Ubuntu:

apt-get install kcachegrind

Get pyprof2calltree, from pip, for example:

pip install pyprof2calltree

Last, run pyprof2calltree with the dump you have copied from the server. It will start KCachegrind automatically:

pyprof2calltree -i 123.124.125.126-2011-01-19\ 18\:49\:35.134598.pro -k

Enjoy

Now you should see your request analyzed in KCachegrind. Find where the problem is and charge your customer dearly.