Re: advice needed: mod_perl reverse proxy

[prev] [thread] [next] [Date index for 2005/04/20]

Stas Bekman wrote:
> allan juul wrote:
> 
>> hi
>>
>> i need advice before i waste too much time on the bleeding obvious.
>>
>> we have a setup where we will reverse proxy content both to our own 
>> backend-servers (which run on IIS) and other external servers which 
>> content we dont control. one of the reasons we proxy is because of 
>> speed/performance
>>
>> we have an Apache 2.054 up front on port 80 and the backend is on the 
>> same machine which is running windows 2004
>>
>>
>> we need to fix broken img src, and absolute links and that sort of 
>> thing coming from the external servers
>>
>> i have fiddled with mod_proxy_html to rewrite stuff and that works ok, 
>> but have some features that doesn't mix well with our solution 
>> (content -type is encoded utf-8, where we proxy to iso-8859-1 for 
>> instance. or some html tags are stripped etc.) also caching becomes 
>> slower because of this output filter it seems (i guess because of 
>> unknown content-length)
> 
> 
> But if you use a mod_perl filter you will still hit the issue of unknown 
> content-length header.

yes, of course that's true.
there goes caching (:

>> it seems way overkill to have a mod_perl enabled frontend, but i'm 
>> pretty confident we could write a mod_perl filter to do the content 
>> rewrites we need.
> 
>  >
> 
>> so, is a mod_perl-enabled Apache acting as a proxy just a sick idea. 
>> it will proxy content and the filter will have to scan all response 
>> content
> 
> 
> It shouldn't be too hard to write a quick prototype, run benchmarks and 
> see whether it scales or not. It's really hard to give an answer when 
> you don't know what kind of type/size of code base you deal with, since 
> if your code base is very small and you don't load tons of other things, 
> you may get away with a quite efficient setup.
> 

i'm not sure what you mean, but the backend code is compiled dll's and 
the usual images/js/css/xsl files all running in a .NET environment. 
it's not a gigantic code base but pretty big. the load on the servers 
are not so bad currently because its spread on several identical machine 
setups (load balanced). its a public web portal.

ok, i'll give it a mod_perl shot. i guess my main worries are just the 
memory the "light" frontend would begin to eat.

btw. (and OT): is it in fact correct that i should see the HTTP headers 
for the proxied server (IIS) and not the frontend (apache) ?



thanks
./allan