Re: acl help

> > ! This cache will not forward your request because it is trying to enforce a
> > sibling relationship. Perhaps the client at
> > ! xxx.xxx.xxx.xxx is a cache which has been misconfigured.
> >
> > Where is my mistake?

> See your miss_access rules. This message is seen when miss_access is
> denied.

We have a few caches that sibling with us. Our siblings are getting alot
of the above errors. Of course, since they are just a peer, they are not
allowed to fetch misses, so get the above error...

The problem appears to be false icp hits.

Here is an output from calamaris that show's the siblings hitting our
squid1.2beta23p4 server

# TCP-Requests Request % kByte % Hit-%
----------------------------------------- ------- ------ -------- ------------
proxy2.omen.com.au 29434 2.76 239840 2.76 78.64
bridge.q-net.net.au 17223 1.61 126434 1.45 72.86

proxy2.omen is squid 1.1, bridge.q-net is 1.2b24

As you can see, there is a 20-30% false icp hit %'age there, hence them
receiving alot of the above errors, due to 20-30% of their requests
failing and them not being permitted to fetch a miss from our cache.

Theoretically, the above two servers should have very close to 100% Hit
rate, since they should only be hitting our server if we answer a HIT to
an ICP request.

By comparison, our squid 1.1 proxy which has several siblings.

# TCP-Requests Request % kByte % Hit-%
----------------------------------------- ------- ------ -------- ------------
proxy.mtx.net.au 18651 2.87 148684 2.93 99.98
obelix.iweb.net.au 8291 1.27 84605 1.67 100.00
caps.esc.net.au 8784 1.35 52732 1.04 100.00

Some of those are squid 1.2's speaking to our 1.1.20. Hardly any false ICP
hits.

And yes, our refresh patterns are synched :)

My question, is there any change in the way squid1.2 handles icp? Our
squid 1.2.beta23 server looks like its saying to our siblings, "Yes I have
this object", and when that sibling tries to grab it from us, we infact
don't have it. I know you can expect a small %'age of false ICP hits, but
20-30% seems a bit high.

Any thoughts on what to do or paths to follow in tracking down the
problem?

Thanks :)

-----
Roger Yerramsetti : ,-------. :
rogery@wantree.com.au : : / / | a n t r e e I n t e r n e t :
Network Administrator : | (_/_/`-'----------------------------- :
Ph (08) 9221 8899 ::::: `-------' http://www.wantree.com.au ::::
Received on Thu Sep 10 1998 - 00:41:31 MDT