On Thu, Jan 17, 2013 at 10:38 AM, Sheng Yang <sheng@yasker.org> wrote:
> On Thu, Jan 17, 2013 at 10:33 AM, Anthony Xu <Xuefei.Xu@citrix.com> wrote:
>> More comments,
>>
>> Can VM access VM by name on IPv6 network( router VM provide DNS service ?)?
>
> Yes, dnsmasq would provide AAAA records.
>
>> Is password-reset service supported on IPv6 network?
>
> Should be in the future, but not phase 1, which only provide DNS and DHCP.
>
>> Is meta-data and user-data service supported on IPv6 network?
>
> Not phase 1.
>
>> Is external network device (F5, SRX) supported on IPv6 network?
>
> Not in the plan.
>
>> What's the impact for Security enabled shared network?
>
> Not in the plan. Only support shared network without SG in the phase 1.
>
>> What's the impact for multiple IPs per NIC?
>
> I guess we may no longer need to have another nic for different public
> subnet, but need to be confirmed.

So I would update the systemvm first, adding the newer version of
dnsmasq and radvd.

Does anyone has specific suggestion on which version to be used? I can
get the dnsmasq from debian testing repo and it works for me. Radvd
can be get from debian stable repo, but I assume it maybe kind of old.

--Sheng
>
> --Sheng
>>
>> Anthony
>>
>>> -----Original Message-----
>>> From: Sheng Yang [mailto:sheng@yasker.org]
>>> Sent: Thursday, January 17, 2013 10:26 AM
>>> To: cloudstack-dev@incubator.apache.org
>>> Subject: Re: [DISCUSS] IPv6 support draft functional spec(phase 1)
>>>
>>> On Thu, Jan 17, 2013 at 10:20 AM, Anthony Xu <Xuefei.Xu@citrix.com>
>>> wrote:
>>> > My misunderstanding, I thought that's the link-local ip in Xenserver
>>> or KVM:-)
>>> >
>>> > If a VM is on both IPv6 and IPv4 network, what's the link-local
>>> address? IPv4? IPv6? Both?
>>>
>>> For dual stack case, we still require IPv6 link-local address only.
>>>
>>> --Sheng
>>> >
>>> >
>>> > Anthony
>>> >
>>> >> -----Original Message-----
>>> >> From: Sheng Yang [mailto:sheng@yasker.org]
>>> >> Sent: Thursday, January 17, 2013 10:13 AM
>>> >> To: cloudstack-dev@incubator.apache.org
>>> >> Subject: Re: [DISCUSS] IPv6 support draft functional spec(phase 1)
>>> >>
>>> >> On Thu, Jan 17, 2013 at 9:49 AM, Anthony Xu <Xuefei.Xu@citrix.com>
>>> >> wrote:
>>> >> > Thanks for the write-up,
>>> >> >
>>> >> > One comment,
>>> >> > Is there any reason not use link-local IPv4 address?
>>> >> >
>>> >> >>*User VM would have one link-local IPv6 address
>>> >>
>>> >> IPv6 required one auto configured link local address per nic(means
>>> >> likely one nic would have more than one IP address, and in the
>>> >> different subnet), and the link local address would be used to send
>>> >> out DHCP request etc(http://tools.ietf.org/html/rfc3315). It's also
>>> >> the basic of Neighbor discovery mechanism in
>>> >> IPv6(http://tools.ietf.org/html/rfc4861).
>>> >>
>>> >> I think IPv4 link-local is less relevant in this case...
>>> >>
>>> >> --Sheng
>>> >> >
>>> >> >> -----Original Message-----
>>> >> >> From: Sheng Yang [mailto:sheng@yasker.org]
>>> >> >> Sent: Wednesday, January 16, 2013 7:11 PM
>>> >> >> To: cloudstack-dev@incubator.apache.org
>>> >> >> Subject: [DISCUSS] IPv6 support draft functional spec(phase 1)
>>> >> >>
>>> >> >> Hi,
>>> >> >>
>>> >> >> The first draft of IPv6 FS is available at
>>> >> >>
>>> https://cwiki.apache.org/confluence/display/CLOUDSTACK/IPv6+support
>>> >> >> now.
>>> >> >>
>>> >> >> Basically based on our previous discussion, we would like to
>>> stick
>>> >> to
>>> >> >> dnsmasq, and assume shared network for advance zone in the phase
>>> one,
>>> >> >> to make thing as simple as possible in phase 1.
>>> >> >>
>>> >> >> Comments/questions are welcome!
>>> >> >>
>>> >> >> --Sheng