Trying to deploy a Django 1.4 project using mod_wsgi and virtualenv, i’m running into a 500. The Apache error_log reveals:
mod_wsgi (pid=30452): Exception occurred processing WSGI script '/path/to/project/site-packages/projectapp/wsgi.py'.
[...] Traceback (most recent call last):
[...] File "/path/to/project/env/myenv/lib/python2.6/site-packages/django/core/handlers/wsgi.py", line 219, in __call__
[...] self.load_middleware()
[...] File "/path/to/project/env/myenv/lib/python2.6/site-packages/django/core/handlers/base.py", line 47, in load_middleware
[...] raise exceptions.ImproperlyConfigured('Error importing middleware %s: "%s"' % (mw_module, e))
[...] ImproperlyConfigured: Error importing middleware projectapp.content.middleware: "cannot import name SomeModel"
From the error message i would expect that this is some kind of a path issue. However, when the offending middleware is removed from the Django settings, the site launches just fine, and there are other modules loaded from projectapp.content, SomeModel is also loaded in this case, otherwise the whole site wouldn’t be able to run.
The import error raised doesn’t come directly from the middleware, as it doesn’t import the model. SomeModel is defined in a speparate app which is actually checked out into the src directory of the virtualenv. But the directory containing this app is also in the path.
The wsgi.py file i’m using:
import os
import sys
sys.stdout = sys.stderr
sys.path.insert(0, '/path/to/project/env/myenv/lib/python2.6/site-packages/')
# The module inside the following directory
# defines SomeModel from the error message
sys.path.insert(0, '/path/to/project/env/myenv/src/some-app/')
sys.path.insert(0, '/path/to/project/site-packages/')
import django.core.handlers.wsgi
os.environ['DJANGO_SETTINGS_MODULE'] = 'projectapp.settings'
application = django.core.handlers.wsgi.WSGIHandler()
Printing sys.path after inserting the module paths shows everything in the expected order, /path/to/project/site-packages/ is listed first and /path/to/project/env/myenv/src/some-app/ (which defines SomeModel) second.
I’ve also tried a different variant, based on the example from the mod_wsgi docs:
import os
import sys
import site
ALLDIRS = [
'/path/to/project/site-packages/',
'/path/to/project/env/myenv/lib/python2.6/site-packages/',
'/path/to/project/env/myenv/src/some-app/',
]
# Remember original sys.path
prev_sys_path = list(sys.path)
sys.stdout = sys.stderr
# Add each new site-packages directory
for directory in ALLDIRS:
site.addsitedir(directory)
# Reorder sys.path so new directories are at the front
new_sys_path = []
for item in list(sys.path):
if item not in prev_sys_path:
new_sys_path.append(item)
sys.path.remove(item)
sys.path[:0] = new_sys_path
#activate_this = '/path/to/project/env/myenv/bin/activate_this.py'
#execfile(activate_this, dict(__file__=activate_this))
import django.core.handlers.wsgi
os.environ['DJANGO_SETTINGS_MODULE'] = 'projectapp.settings'
application = django.core.handlers.wsgi.WSGIHandler()
The error logged by Apache is exactly the same.
I have to add that the middleware is loaded just fine when i’m running the development server on the same machine, so i have no idea what’s going wrong. Is there a way to get a better traceback from mod_wsgi?
EDIT:
Fogot to mention that i’m using mod_wsgi in daemon mode. Here are the relevant parts from my vhost config:
<VirtualHost x.x.x.x:80>
WSGIDaemonProcess foo user=foo threads=10 umask=0002
WSGIProcessGroup foo
WSGIScriptAlias / /path/to/project/site-packages/projectapp/wsgi.py
</VirtualHost>
Okay after hours of debugging this turned out to be a race condition caused by the middleware. The middleware i’m using is similiar to the
FlatpageFallbackMiddlewarefrom Django contrib and actually the import of the view caused the problem.Moving the import statement inside the
process_responsemethod solved the problem for me: