Have verified that cache_peer_domain, cache_peer_access and
neighbor_type_domain can be used to only forward !.com to parent. Actually
we could have use hierarchy_stoplist .com too, but, this will also include
.com.x traffic.
I notice that putting proxy-only in cache_peer (not to cache !.com
objects) will somehow cause the .com objects not to be cached also. Did i
miss out something in the config?
cache_peer notdotcom.box parent 8080 3130 proxy-only no-query no-digest
hierarchy_stoplist cgi-bin ?
acl QUERY urlpath_regex cgi-bin \?
no_cache deny QUERY
acl COM url_regex ^http://.*\.com(/|\b)
cache_peer_access notdotcom.box deny COM
Rgds,
Wei Keong
On Thu, 18 Jul 2002, Wei Keong wrote:
> Seems that cache_peer_access and cache_peer_domain send UDP query (ICP?)
> to peers. This might cause some congestion in heavy traffic...
>
> If I want to do 'blind' forwarding based on domain, is there anyway to
> disable the UDP query, like no-query in cache_peer?
>
> How about neighbor_type_domain? Does it also send UDP query to peers?
>
> Rgds,
> Wei Keong
>
>
> On Wed, 17 Jul 2002, Henrik Nordstrom wrote:
>
> > Wei Keong wrote:
> > > > cache_peer the.notdotcom.box parent 3128 0 no-query connect-timeout=2
> > > > proxy-only
> > > >
> > > > acl COM dstdomain .com
> > > >
> > > > cache_peer_access the.notdotcom.box deny COM
> > > >
> > > > prefer_direct off
> > >
> > > We want the .com box to handle all .com traffic and only forward
> > > the rest to nondotcom box. Shouldn't we use sibling instead of parent?
> >
> > No. You want to forward all notdotcom, not just what the other already has in
> > it's cache.
> >
> > > Actually, to differentiate .com and !.com, we can also use
> > > cache_peer_domain. To avoid reverse dnslookup, it will be better not to
> > > use dstdomain, right?
> >
> > Right, unless you want those reverse dnslookups.. you can also use a url_regex
> > to match IP addresses if you want to avoid this but still use dstdomain acls.
> >
> > Note: Squid should probably be extended with a sane ACL method to determine if
> > a URL is specified using a hostname or a IP address... writing a correct
> > url_regex pattern is not trivial.
> >
> > Regards
> > Henrik
> >
>
>
Received on Thu Jul 18 2002 - 00:52:21 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:09:17 MST