Re: [squid-users] Squid 2.5STABLE3 Questions

On Wed, 23 Jul 2003, Henrik Nordstrom wrote:

> ons 2003-07-23 klockan 15.42 skrev Merton Campbell Crockett:
>
> > Question: Is there any good documentation on the interaction between the
> > cache_peer, cache_peer_domain, neighbor_type_domain, always_direct, and
> > never_direct configuration directives. There seems to have been significant
> > changes in this area from earlier versions of Squid and that it is now more
> > difficult to construct a viable Squid hierarchy.
>
> No major changes here for a very long time..
>
> I would recommend ignoring the cache_peer_domain directive. The
> cache_peer_access directive performs the same function and gives better
> control.
>
> neighbor_type_domain is rarely if ever needed so you can most likely
> ignore this as well. You will know if you need this one.
>
> > In preliminary testing, I've had a cache_peer defined as a parent return an
> > error status rather than the requested content. This behaviour presents a
> > major problem for interior network enclaves that can never directly access
> > external web content.
>
> What error?

  Valid document was not found in the cache and only-if-cached was specified

This error is returned under the following conditions.

   (1) The browser requesting the document is configured to use a Squid
        Cacheing Proxy that is running Squid 2.5STABLE3
   (2) The requested document is served by an Apache server running on
        the same system as a Squid Cacheing Proxy.

The first instance in which I noticed this the request had been forwarded to
a cache that was a parent of the initial cache. My expectation was that the
parent cache would retrieve the document; however, the parent cache happened
to have a peer cache running on the same system as the web server providing
the content.

The document could be retrieved by setting allow-miss for the parent in the
first proxy.

The next instance in which this occurred the content being retrieved was
served from an Apache server running on the same system as a peer of the
first proxy.

I don't recall seeing this type of behaviour prior to using 2.5STABLE3.

This leads to a rather interesting problem. My intent is to use Squid to
manage traffic between enclaves in an overlay network. I would like to have
Squid query proxies at adjacent enclaves and along the path to the proxy at
the Internet gateway node so that cacheable data does not have to traverse
the entire overlay network.

What I would want to happen is that regardless of the only-if-cached setting
that Squid retrieve and deliver content from its "local" system for some
definition of "local".

Merton Campbell Crockett

-- 
BEGIN:				vcard
VERSION:			3.0
FN:				Merton Campbell Crockett
ORG:				General Dynamics Advanced Information Systems;
				Intelligence and Exploitation Systems
N:				Crockett;Merton;Campbell
EMAIL;TYPE=internet:		mcc@CATO.GD-AIS.COM
TEL;TYPE=work,voice,msg,pref:	+1(805)497-5045
TEL;TYPE=fax,work:		+1(805)497-5050
TEL;TYPE=cell,voice,msg:	+1(805)377-6762
END:				vcard