Re: [squid-users] want squid 3.1.10 to stop giving IPv6 errors fetching www.google​.com

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Sun, 19 Aug 2012 18:23:20 +1200

On 19/08/2012 4:06 a.m., Will Roberts wrote:
> On 08/18/2012 08:02 AM, Robert Collins wrote:
>> On Sat, Aug 18, 2012 at 2:51 PM, Bennett Haselton
>> <bennett_at_peacefire.org> wrote:
>>> I installed squid 3.1.10 on CentOS 6.3 with the default squid.conf.
>>> When I test it out from localhost:
>>>
>>> The following error was encountered while trying to retrieve the URL:
>>> http://www.google.com/
>>> Connection to 2607:f8b0:4004:800::1014 failed.
>>> The system returned: (110) Connection timed out
>>
>> If you have IPv6 configured on your machine but don't have IPv6
>> connectivity to the rest of the world, I would expect this symptom.
>> Solutions:
>> - don't have IPv6 configured
>> - or have connectivity.
>
> Sometimes that's not fully in your control. One machine I have doesn't
> have a global IPv6 address, or any IPv6 routes, yet still tries to use
> IPv6 addresses when reaching servers. I'll admit I spent no time
> investigating as I already knew how to force squid to use IPv4:
>
> tcp_outgoing_address <ipv4 addr> <acl>
>
> I have multiple outgoing IPs, so mine are usually of the form:
>
> acl out0 myport 80
>
> tcp_outgoing_address <ip> out0
>
>
> --Will

There is no reason this *timeout* should occur on a properly configured
network. ICMPv6 messages from the closest router or even from the kernel
internal NIC subsystems should be triggering a "unable to connect" error
instead which in turn triggers failover to the next IP address in under
a millisecond.

3.1 series have a few problems in how many failovers are done, and using
some IPs multiple times. But these issues are resolved in the 3.2
series. Please upgrade.

Amos
Received on Sun Aug 19 2012 - 06:23:28 MDT

This archive was generated by hypermail 2.2.0 : Sun Aug 19 2012 - 12:00:03 MDT