14

I am currently running a Centos 6.4 server, with Apache 2.2.15 and mod_wsgi 3.2. The server is hosting a django-based site (django 1.5.1, python 2.6.6). Everything was running fine until I installed scipy 0.12.0 via pip. Now, when I attempt to load the django app, the server does not respond, and it appears that child httpd processes that are spawned hang. Looking through my logs (/var/logs/httpd/error_log, my vhost error.log, and my system logs) yield no errors.

If I load my models, etc.. via the django manage.py shell, everything works fine, which leads me to believe it is a mod_wsgi issue.

Any thoughts on how to start troubleshooting this?

MarkD
  • 243
  • 1
  • 2
  • 5

2 Answers2

33

Some third party packages for Python which use C extension modules, and this includes scipy and numpy, will only work in the Python main interpreter and cannot be used in sub interpreters as mod_wsgi by default uses. The result can be thread deadlock, incorrect behaviour or processes crashes. These is detailed in:

http://code.google.com/p/modwsgi/wiki/ApplicationIssues#Python_Simplified_GIL_State_API

The workaround is to force the WSGI application to run in the main interpreter of the process using:

WSGIApplicationGroup %{GLOBAL}

If running multiple WSGI applications on same server, you would want to start investigating using daemon mode because some frameworks don't allow multiple instances to run in same interpreter. This is the case with Django. Thus use daemon mode so each is in its own process and force each to run in main interpreter of their respective daemon mode process groups.

Graham Dumpleton
  • 5,990
  • 2
  • 20
  • 19
  • Hi Graham, could you update this answer in the context of more recent versions of mod-wsgi? Specifically, is this supposed to be a problem by default if I have configured apache using mod_wsgi-express? In the generated `httpd.conf` file, `WSGIApplicationGroup` is not used. However, there is `application-group=${GLOBAL}` in the `` and `` blocks. I see a WSGIDaemonProcess directive in the generated `httpd.conf` file. Does that mean it's already using daemon mode by default? – Kal Dec 07 '17 at 09:12
  • If you use ``mod_wsgi-express start-server`` or Django integration for mod_wsgi-express it runs with daemon mode as default and use the main interpreter. So this is not a problem in that case. If you manually configure Apache, then is still an issue. The ``ONE_PROCESS`` part is only for when you force it into debug mode, in which case it runs in single process embedded mode. It still runs in main interpreter though. – Graham Dumpleton Dec 07 '17 at 10:54
  • The ``application-group`` option on ``WSGIScriptAlias`` is an alternative to using ``WSGIApplicationGroup``. – Graham Dumpleton Dec 07 '17 at 10:55
  • I had this error come in random (probably because of numpy) and crash my site. I just did what you said and I don't know if it works. Do you have any idea on how I could try to reproduce the error ? – Sam May 04 '20 at 10:59
4

Another solution that fit my way of configuring WSGI was changing the WSGIScriptAlias line:

WSGIDaemonProcess website user=user group=group python-path=/path/to/venv/website:/path/to/venv/lib/python2.7/site-packages
WSGIScriptAlias /website /path/to/venv/website/wsgi.py process-group=website application-group=%{GLOBAL}

<Location /website>
        WSGIProcessGroup website
</Location>

<Directory /path/to/venv/website>
        WSGIScriptReloading On
        <Files wsgi.py>
                Allow from all
                Require all granted
        </Files>
</Directory>

note the attributes

process-group=website application-group=%{GLOBAL}

which are usually not required

Nils Werner
  • 141
  • 2
  • 1
    You can drop WSGIScriptReloading directive as that defaults to on and generally would never need to be turned off. Due to using process-group option to WSGIScriptAlias, you can also drop the WSGIProcessGroup directive. – Graham Dumpleton Jun 08 '15 at 23:56