WCCPv2 doesn't specifically "work like that" but it doesn't specifically not "work like that." The primary cache programs the router(s) with maps for distributing traffic. The current setup tries to balance traffic based on the bits (port, ip, etc) that are configured and the weight influences the hash/mask assignment. You -can- program the distribution maps -any way you want-, so you could write a custom distribution map to map 100% of traffic to cache A. Cache B can then advertise itself as the 100% traffic map when A fails. The way -that- works is tricky - the cache notices that the primary cache isn't sending it messages and it elects a new master. This can take uhm, 2 or 3 WCCP2_I_SEE_YOU intervals and thus it isn't fast. The alternative there is to have the caches ping each other and when cache X disappears, try to force a re-election somehow. I forget if thats possible - I know that the cache itself can remove itselfr from the WCCPv2 service group but I don't know if an election can be forced from another cache. HTH, Adrian On Thu, May 22, 2008, Nick Duda wrote: > I've deployed my new solution and its working well, but see some issues realted to transparent authentication (I'm working with SecureComputing on this). I have one question though that would help solve this issue. > > Currently, I have 2 squid caches configured as wccp with a router. In viewing the logs it shows that the proxies are being used randomly (in a load balance way) with the router, which is cool but im seeing some issues with SmartFilter transparency doing that. Is there a way through wccp config in squid or on the cisco gear to only use one cache until it goes down then use the other? > > All WCCP requests should go to one of the 2 configured wccp enabled squid caches until it goes down, then then other takes the requests. > > Thanks, > Nick -- - Xenion - http://www.xenion.com.au/ - VPS Hosting - Commercial Squid Support - - $25/pm entry-level VPSes w/ capped bandwidth charges available in WA -
