Re: WCCP glitch with 2.3S1

From: Garlic <garlic@dont-contact.us>
Date: Wed, 26 Jan 2000 11:43:25 -0800

First, let me say I am very happy with the WCCP implementation. Thanks for
the hard work you put in on this.

To me a glitch is not something that is broken. Its just an oddity that must
be taken into consideration when connecting things. Perhaps a better word
would be limitation?

I would like to suggest that an allowed list of WCCP hosts be added to the
implementation. I can foresee a situation where the route between the cache
and the router changes due to a failure. Another condition would be where
you are doing load balancing across different routes between the cache and
the router.

Of course, a better solution would be for Cisco to have a configuration
parameter to set the WCCP address for the router so that all packets are sent
from that address.

Roy

Glenn Chisholm wrote:

> > Interesting little glitch I found while installing WCCP and Squid
> > V2.3S1.
> >
> > The cache must point to the primary IP address on the router on the
> > interface that the server is attached/routed thru. A secondary address
> > will not work The symptoms are that the router sees the cache but the
> > cache state never shows useable and the connection times out after 30
> > seconds and resets.
> >
> > I suspect that the source address of the WCCP packet is the primary
> > address and squid ignores it.
> >
>
> It is not a glitch, it is deliberate. I wrote it that way so that the
> proxy would only accept packets from the host that you specified in the
> squid.conf. If it does not do that it leaves you very open.
>
> glenn
Received on Wed Jan 26 2000 - 12:53:31 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:50:43 MST