Re: Ranting on ipcache bogosity

Perhaps if ipcache stored all the addresses for a host and return
them all to Squid when asked, perhaps using the same method as BIND.
 
Then Squid could try all the IP addresses returned by ipcache. This
would still not fix the "black hole" problem, but I suspect that's a
more general problem with the host than an implementation problems
with Squid.

I must say that while Squid does have a few problems, it's far, far
ahead of any other cache software that I've seen.

--Eric

Some time ago Chris Fedde said:
>
> The Problem:
>
> With positive_dns_ttl set to zero squid exhibits a behavior which
> is confusing for multi-server sites with inconsistent failure modes.
> The behavior with positive_dns_ttl set to non zero is much worse so I
> will not even go into that (see the ps below).

[snip]

> My Recommendation:
>
> While the current behavior is to be expected given the ipcache
> implementation. I suspect that a better behavior would be easy
> to implement. I'd like to see squid keep information about the its
> successful connection attempts. The little discussion below details
> my thinking on these lines.
>
> 1) Forget about caching DNS results in the server. Named
> does a fine job of it and the async technique used by the
> dnsservers work well to access a local (or remote) IP
> address cache. Further, the current ip cache implementation
> is broken.
>
> 2) Cache _failed_ ip addresses for each HTTP attempt. Set
> a TTL for these cache entries at something like the TTL
> for the DNS record itself.
>
> 3) Check the failed ip cache before making a connection attempt.
>
> 4) Try ALL THE ADDRESSES returned by DNS before returning a failure
> to the client.

-- 
Eric Wieling
Advanced Network Research
InterCommerce Corporation
Pager: 800-758-3680
The world needs no help seeing a fool for what they are.