Re: Q: Parent running but not caching?

scoanc@thor.cf.ac.uk writes:

>On Fri, 9 Aug 1996, Peter Childs wrote:
>
>> In article <4uci21$4pl@al.imforei.apana.org.au> you wrote:
>>
>> : Is there a way to configure squid to cope with a Netscape parent which is
>> : up, but not running the cache software ? At the moment squid gets a reply
>> : to its UDP enquiry (bounced off port 7) because the parent is running
>> : TCP/IP, but gets a rude reply when it tries to make a TCP connection to
>> : the data port (8080).
>>
>> If the server doesn't support caching/proxying then don't have it as
>> a parent.

This is hard. You might want to run a little Perl process which
simply echos UDP packets. If you could make this process start
and stop at the same time as Netscape, then you get the same
functionality.

>In this case the server is supposed to be running a cache, but isn't. I
>haven't been able to find out whether the cache program has been stopped
>for development purposes or has broken. The problem is that squid's check
>for aliveness (against a Netscape cache) only checks that UDP on the
>machine is alive, not the cache software. Since the inactive parent always
>responds fastest, squid is helpless until I manually remove the offending
>parent from the tables. A couple of possible solutions come to mind (there
>are no doubt others):
>
>1) Check aliveness of the cache s/w, not the machine (does the Netscape
> cache make any visible difference to the machine it runs on?)
>2) Squid could perhaps record the fact that a host is returning 'connection
> refused' on it's supposed cache service port, and not use the parent
> again until some timeout expires. The individual request which
> detected the problems could either be failed, or re-processed with the
> modified list of parents.

Actually we do #2 already. Any neighbor which gives us a failed
connect() will not be used for the following 60 seconds.

Duane W.
Received on Wed Aug 14 1996 - 13:30:50 MDT