On 07.06.2012 17:19, Amos Jeffries wrote:
On 7/06/2012 10:48 p.m., Eugene M. Zheganin wrote:
Hi.
Happy world ipv6 launch day to everybody. :P
squid/3.1.12, FreeBSD 8.x|9.x
a) why when using ipv6 address, like fd00::316 or, no matter
[fd00::316], squid detects that configuration file is corrent, but
then I'm immidiately getting "TCP connection to fd00::316/3128
failed", but, in the same time, I CAN telnet to the fd00::316 port
3128 ?
More to the point *what* is failing to connect to your Squid and from
where?
Assume I get one parent squid A, and one child squid, B.
B's IP (v6) is in the parent's config file, like
acl children src fd00::d02/128
B's config has the line like
cache_peer fd00::316 3128 3130 default
or
cache_peer [fd00:316] 3128 3130 default
and some lines like
always_direct deny some-domain-based-acl
always_direct allow all
never_direct allow some-domain-based-acl
never_direct deny all
In this situation I'm starting to get these lines on B:
TCP connection to fd00::316/3128 failed
In the same time I can connect from B's console (telnet fd00::316 3128).
b) more long story. I think I'm just not getting something. From the
2.6 version I always had to use no-query option for parents. That is
because the cold
ZOMG. Did I write this ? I mean 'children'.
squid is always detecting the parent proxy as dead, but the weird
thing is, it's still capable of communicating to them. I mean I see
udp/3130 packets in tcpdump going both ways, and UDP_HIT/000
UDP_MISS/000 ICP queries from children on parent proxies.
Why is that ?
bugs?
That's too easy to explain. I think the situation when I discover and
report this bug first time for like 10 years is quite impossible.
Eugene.
