Re: transparent cisco configuration

In message <Pine.GSO.3.96.980402205045.24575G-100000@dinosaur>, Ming Lu writes:
>Not true. You can set route-map and set netx-hop with mutiple ip
>addresses; i.e., if first one fails, then the http traffic will go to the
>second one, etc.., but once all of squid servers fail, you customers' http
>traffic will be sucked into a "blachole" with no return. Any cisco cureent
>stable release IOS support this.

nope. this was *exactly* the point i was attempting to make.

a route-map is only looking at things in terms of a per-packet perspective.
if the _host_ running the cache is fine, but the cache itself is dead or
not responding, you *won't* get the failover.

you will *only* get fallover if the link-layer of the network interface
is down on the router.

>WCCP is total different story. It deos very nice job on loadbalancing,
>failover traffic, very easy to configure. Cisco really know how to handle
>traffic on the routers.

yes. shame WCCP isn't documented as a rfc.

wccp has heartbeats from the CCE's to the wccp-router, saying "as a proxy,
i'm still functioning". if a timeperiod goes by without heartbeats, then
the router stops forwarding http connections to that CCE.
as far as the other issues of loadbalancing and CCE internals, i don't
think that the squid list is the forum for discussing how it works.

cheers,

lincoln.
Received on Thu Apr 02 1998 - 18:11:46 MST