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@xxxxxxxxxxxxx> 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
