> On 14 พ.ค. 2558, at 05:26, Tom Evans <[email protected]> wrote: > >> On Thu, May 14, 2015 at 4:36 AM, reduxionist <[email protected]> >> wrote: <snip> >> The reason for those search results is that mpm-prefork does, however, spawn >> a process per request, > > No, really, it does not. It only spawns a new process when there are > no available workers to process an incoming request, and you have not > reached the maximum number of workers that you have configured it to > start.
Oops, sorry, I was totally wrong about that. > You can configure it to start all the worker processes you want > when it starts up, and never to kill them off, and it will never spawn > a new process. > > Apache processes are small, unless you do daft things like embed your > web application in each worker process (mod_php style). This is the > main complaint "Apache is eating all my memory" - it isn't, your web > application you've embedded into Apache is eating all your memory. > > All of this is irrelevant for django, because with Apache you should > use mod_wsgi in daemon mode, which separates out your web application > processes from the web server. <snip> >> I think the reason a lot of people seem to run mpm-prefork is just that it's >> the default multi-processing module for Apache on most (all?) *nix platforms >> and they don't know any better. > > Quite. We run a pair of Apache 2.4 reverse proxies in front of all of > our (400+) domains, serving around 40 million requests per day, > providing SSL termination and static file serving. We use event MPM > and we have it scaled to support a peak of 2048 simultaneous > connections. Load on the server never goes above 0.2, memory usage > never goes above 1GB for the entire OS + applications, the rest of the > RAM is used by the OS to cache the aforementioned static files. > > On our app servers we typically use Apache with worker MPM and > mod_wsgi, although we have a few nginx+uwsgi sites, and I would dearly > love some time to play around with a circusd + chausette + celery > setup. > > The choice of web server is, these days, irrelevant. If it uses too > much memory or can't handle enough users, it is never the fault of the > web server, but instead of your application and/or configuration. Thanks for the awesome write-up Tom and for correcting my error, my apologies to all for contributing to any FUD! Yours gratefully, Jonathan -- You received this message because you are subscribed to the Google Groups "Django users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/django-users. To view this discussion on the web visit https://groups.google.com/d/msgid/django-users/46CB380A-BBA2-4F78-8F7F-8DFC7A994859%40gmail.com. For more options, visit https://groups.google.com/d/optout.
