[Pkg-haproxy-maintainers] Bug#726323: Bug#726323: Bug#726323: haproxy doesn't start on boot due to missing IPv6 address on interface

[Philipp Kolmann]

On 2013-10-14 23:52, Apollon Oikonomopoulos wrote:
> On 19:54 Mon 14 Oct     , Vincent Bernat wrote:
>>   ❦ 14 octobre 2013 16:07 CEST, Philipp Kolmann <philipp at kolmann.at> :
>>> Also I have changed to init.d file to source a second config file,
>>> where I have
>>> the config which both nodes share, so I can easily copy one file.
>>> If you need the config of the global file, please tell me.
>> Well, there is little we can do. I suppose the address is setup in
>> network/interfaces but not available until it passes the DAD check. With
>> IPv4, you could circumvent the problem by setting
>> net.ipv4.ip_nonlocal_bind but there is no such setting for IPv6 and
>> kernel maintainers do not want to have such a setting.
> I was thinking either that, or the address is managed by keepalived and
> subject to all sorts of master election delays. In either case,
> additionally assigning the address to a dummy interface could make
> things work (dummy interfaces do not perform DAD).

The IPv6 addresses which are balanced with keepalived are all configured 
with transparent. There I don't have any issues.

I have the issue with the fixed IPv6 Address of the interface, specified 
in /etc/network/interfaces

iface eth0 inet6 static
         address 2001:629:1005:30::11
         netmask 64


This address takes some seconds to become available.

And I have in my global file following peers:

peers ha-cluster-v6
         peer ha1 2001:629:1005:30::11:1024
         peer ha2 2001:629:1005:30::12:1024


and here haproxy tries to bind the port to the IPv6 address but it isn't 
available.

Thanks
Philipp