Re: [squid-users] Cookie problem

>>>>> On Thu, 31 May 2007 15:49:05 +0200
>>>>> henrik@henriknordstrom.net(Henrik Nordstrom) said:
>
>
> --=-hQpkNJlx1qzrQqzbkcT7
> Content-Type: text/plain
> Content-Transfer-Encoding: quoted-printable
>
> tor 2007-05-31 klockan 13:39 +0200 skrev Ludovit Koren:
>
> > just last question. Is it planned to work correctly the cache
> > hierarchy I mentioned earlier any time soon?
>
> Which was? (reading some hundreds email a day...)
>
> Regards
> Henrik
>
>

my question with your answer follows:

>>>>> On Mon, 28 May 2007 22:12:10 +0200
>>>>> henrik@henriknordstrom.net(Henrik Nordstrom) said:
>
>
> --=-d9Fl78gvdQZD+JvJS9o/
> Content-Type: text/plain; charset=utf-8
> Content-Transfer-Encoding: quoted-printable
>
> m=C3=A5n 2007-05-28 klockan 14:57 +0200 skrev Ludovit Koren:
>
> > The caches work ok. The problem is when a client requests a page where
> > a cookie is set. When all the pages are served using the same parent
> > cache it works correctly. When the parent cache changes I get an error
> > message that the cookie is not valid (it is coming from another
> > IP). It does not matter if I am using any options on the cache_peer
> > configuration line (I used round-robin and sourcehash).
>
> The only forwarding mechanism where this works is sourcehash, but you
> must then disable everything else (ICP, Digest, HTCP, ICMP), and parent
> cache hit ratio will suffer.
>
> As a mid-approach you can build a blacklist of broken sites which
> assumes the IP will remain unchanged for a complete browsing session and
> send these to a set of cache_peer using sourcehash.. (might be the same
> peers but with different names..). See cache_peer_access for how to
> select what gets sent to which peers, and the name option to cache_peer
> for specifying the same peer multiple times with different options.
>
> Regards
> Henrik
>

Regards,

lk
Received on Fri Jun 01 2007 - 01:40:43 MDT