Re: Forwarding loops with squid 2.0 RELEASE

On Tue, 20 Oct 1998, Niall Doherty wrote:

> > > Try Squid-2.0.PATCH2. It has fixes for digest forwarding loop bugs.
> >
> > I have now upgraded three of our external caches to Squid-2.0.PATCH2.
> > There are all set up as mutual peers.
> >
> > The incidence of the loop warnings has certainly diminished, but it has
> > not disappeared:
>
> The fix Duane is referring to breaks any forwarding loops that are
> detected. It doesn't stop them occurring.

Given the frequency, I suspect they are only happening on false hits.
This is logical, at least. True hits won't generate a loop.

> One way of stopping them is to configure your cache so that any
> requests that originate from siblings, do not cause requests to
> go out to any other siblings (and the other sibling maintainers
> should do the same). Here's an example:
>
> # siblings you know about
> acl from_peers srcdomain "/usr/local/squid/etc/trusted_caches"
>
> # caches you use as *siblings* only - parents not necessary
> cache_peer_access proxy1.me.com deny from_peers
> cache_peer_access proxy2.me.com deny from_peers
>
> Cheers,
> Niall

I would have thought it more suitable for this to be the default
behaviour. That way if I configure my cache in the naiive fashion, it
works without complaint.

I can't think of a good reason why anyone would want requests from
siblings to be resolved through other siblings. And especially not from
the originating sibling.

John

John Sloan
UK Unix Systems Administrator
UUNET Worldcom
Received on Tue Oct 20 1998 - 08:04:54 MDT