> > There's also a talk I gave a couple of years ago about non-plack > options, like nginx+fastcgi: > http://elem.com/~perrin/web_arch.pdf.gz >
Thanks for posting this. I understand the general idea of having a lightweight server on port 80 to handle static content and reverse proxying to another server on the back end for dynamic content, but have difficulty making this work in practice. Basically I want to have various virtual hosts be reverse proxied to various back end servers (e.g. mod_perl for some legacy apps, a Catalyst app under fast CGI). I tried this with Nginx and it all seemed to work but what I discovered is that over time the backend Apache processes were using more and more memory, whereas when I go back to a monolithic Apache the memory usage stays pretty stable. In your slideshow you talk about using Apache with a threaded worker MPM as the frontend and a prefork MPM as the backend. So I'm trying to set that up with Apache 2.2.22 but I'm having problems serving basic content (I'm getting text but not formatted HTML) Is there a recommended version of Apache for doing this? I've been googling around for detailed info on how to do this and found a few things, but any links or suggestions are greatly appreciated. Dan