On Wed, Jul 13, 2011 at 1:20 PM, Amos Jeffries <squ...@treenet.co.nz> wrote: > Thank you. > > Because the way chained proxies work proxy1 is just another client to > proxy2. > > Are they both caching? that would make proxy1 only pass requests through > proxy2 when the object is expired/stale. In that case both proxies will > detect it as a MISS at once, but HIT only get as far as the first proxy. > proxy2 can only be a HIT if some source other than proxy1 caused that > object to be cached or updated before proxy1 had a MISS/REFRESH on it. > > So... > Two proxies chained like this only really make sense if one is not caching > (proxy2 usually for efficiency), or if there are multiple proxies/clients > plugged into router2, or if the cache space available to proxy1 is smaller > than the one able to be stored in proxy2. > > > Also, proxy1 can use cache_peer directive to pass requests through proxy2. > There is no need for it to be transparent/intercepting at any level other > than the one directly next to the client. > > Amos > -- > Please be using > Current Stable Squid 2.7.STABLE9 or 3.1.14 > Beta testers wanted for 3.2.0.9 >
First, they both caching. Client behind proxy1 gets HIT object when proxy1 done caching any object. And so is the client behind proxy2 gets HIT object whenever proxy2 done caching any object. the weird thing is, proxy1 won't get any cache from proxy2. both proxies has the same RAM, configuration, OS and cache space. About cache_peer directive, 1. would this mean, proxy1 gets caches from proxy2? or proxy1 gets direct request to internet, without checking caches on proxy2 first? 2. What protocol this would be? Does zph_mode tos work here? So i can set queue limit on my router based on HIT or MISS objects. Thanks for the detailed explanations :)
