Have we been opted out of IPv6 AAAA resolution?

Phil Mayers p.mayers at imperial.ac.uk
Wed Jun 6 16:47:43 CEST 2012


On 06/06/12 15:23, Shumon Huque wrote:

> Although I didn't mention it in my earlier e-mail, we did have a
> fiber cut induced outage to our external IPv6 connectivity for
> many hours on Tuesday, May 29th. I did mention this to dns-admin@
> google.com and they confirmed that this long outage skewed their
> IPv6 brokenness measurements and put us on the blacklist, and
> that it will likely take a couple more days to age out of the list.

That's interesting. Google have confirmed that we have been blacklisted 
as a result of a sudden increase in brokenness, starting on... the 29th 
as well! This is really odd - as far as we're aware, nothing broke at 
our end on that day.

Mysteriously, the brokenness apparently dropped off on the 31st, and 
then came back the following day. Again, as far as we can tell, nothing 
changed locally on these days.

> So, Google is actively measuring IPv6 connectivity issues and
> dynamically maintaining a AAAA blacklist. Not sure if any of the
> others are, so far we seem to be getting AAAA from everyone else.

We have been pointed at this:

http://www.google.com/intl/en_ALL/ipv6/statistics/data/no_aaaa.txt

...which is linked from here:

http://googleblog.blogspot.co.uk/2012/06/world-ipv6-launch-keeping-internet.html

I'm currently writing a script to process our outbound URL logging port 
mirror and spot brokenness; the key seems to be URL callouts to:

p<randomstring>.<number>.{i1,i2,s1}.{v4,ds}.ipv6-exp.l.google.com

I have asked Google if they can provide info on how all this works.

I'm a bit puzzled as to where our apparent 1.5% brokenness ratio is 
coming from; it seems huge, and I'm concerned it might be some slowness 
in IPv6 NS/ND processing (as per my other post on cisco-nsp recently, 
for those who read there).



More information about the ipv6-ops mailing list