You have to also comment out the LoadModule line for mod_python, with
full stop/stop of Apache after that.
This may well not be the issue, but mixing mod_python and mod_wsgi
where they don't both link to the Python shared library is a know
source of crashes. So, rather eliminate this as issue first. We can
then look at what application you are running and what third party
modules for database etc you are using.
BTW, does the Apache instance also load PHP module?
Graham
What command did you use to stop/start Apache?
This looks like you did not do a full stop and then a start, but just
a reload or restart. In other words, it didn't actually shutdown the
parent Apache process completely, but just told it to try and reload
stuff within same process.
>> BTW, does the Apache instance also load PHP module?
>
> Yes. I am also using MySQL with PHP and I installed the MySQL python
> bindings a while back (they are probably using the static python
> library). I'm also running SVN.
Use of PHP in same Apache can also cause issues, but in general if it
is going to cause a problem it will do it in embedded mode as well as
daemon mode.
Still not happy at this point that Apache was completely stopped, so
double check that it was and post commands you are using to do it.
Graham
Can you ensure you have set:
LogLevel info
instead of default of:
LogLevel warn
that Apache generally uses.
Then look at the Apache error logs and see if you can confirm whether
the processes that are crashing are the mod_wsgi daemon processes. If
you can post an example of messages from logs for same pid that shows
where in mod_wsgi initialisation of daemon process it got, that would
help.
Also, do you have SELinux extensions enabled for the operating system?
These have been known to cause daemon mode processes to crash on
startup or otherwise not work before.
Graham
It is possible you are encountering the following issue:
http://code.google.com/p/modwsgi/issues/detail?id=40
If User/Group directives in Apache configuration files don't proceed
WSGIDaemonProcess and you haven't set user/group for
WSGIDaemonProcess, it will crash when starting up daemon process. Was
thinking whether this was issue before, but wanted to eliminate
mod_python as cause first.
So, are the User/Group directives set in Apache configuration and if
so to what? Are those directives set before WSGIDaemonProcess would be
read as part of configuration?
Does the problem go away if you supply user and group options to
WSGIDaemonProcess where the user and group is set to what is
appropriate for your own account?
WSGIDaemonProcess mysite.com processes=2 threads=15 user=yourusername
group=yourdefaultgroup display-name=%{GROUP}
Graham
My understanding is that they should always be set as there is no
guarantee that the defaults of -1/-1 for each would match to a valid
user in the passwd/groups file of a particular operating system. If
the Apache is a binary distribution from Fedora and they have
commented these out in default configuration, that is most odd.
Is Apache a binary distribution or self built from source code and you
had commented out User/Group yourself?
Graham
The way the web host has set up order of inclusions, if using config
snippets, is probably then that the User/Group are not set early
enough and WSGI stuff is getting processed before they have been set.
User/Group is really one of those things that should be set early on,
as modules could rely on what the values have been set to, but then
modules could also delay looking up values until after configuration
phase. The latter is what mod_wsgi really needs to do, but haven't got
around to it.
Graham