incubator-cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sheng Yang <sh...@yasker.org>
Subject Re: [ASFCS41] IPv6 Support
Date Wed, 16 Jan 2013 02:10:34 GMT
On Mon, Jan 14, 2013 at 3:34 PM, Sheng Yang <sheng@yasker.org> wrote:
> On Mon, Jan 14, 2013 at 10:51 AM, Sheng Yang <sheng@yasker.org> wrote:
>>
>> On Mon, Jan 14, 2013 at 10:45 AM, Chiradeep Vittal
>> <Chiradeep.Vittal@citrix.com> wrote:
>>>
>>> You can if you assume that it is of LL-type. Windows requires a registry
>>> fix to generate an 'LL' type DUID.
>>
>>
>> Does dhcp6c support it? Didn't see a word from the document about DUID type,
>> and it genereate LLT type by default Probably I should try another dhcpv6
>> client.
>>>
>>>
>>> You can also get dnsmasq to ignore the duid and use the mac.
>>
>>
>> Haven't confirmed it's possible to use the mac, asking dnsmasq mailing
>> list...
>
> Get answer from Dnsmasq's author Simon, it's impossible to use MAC to
> identify client right now, since there is no such dhcpv6 standard.
>
> So it looks like we need stick to DUID-LL.

I am failed to generated DUID-LL from the latest version of dhcp6c in
CentOS 5.6. Is DUID-LL poorly supported by dhcpv6 client? Any other
suggestion?

User may need to install new dhcpv6 client to use with CloudStack,
that sounds not that good...

I am working on the FS right now, since we have the key steps(except
DUID-LL, which should be a dhcpv6 client issue), and likely able to
get it done tomorrow.

Since we're working on advance shared network, it's easier to deal
with createNetwork API rather than createVlanAndIpRange API.

--Sheng
>
> --Sheng
>
>
>>
>> --Sheng
>>
>>
>>>
>>>
>>> On 1/14/13 9:54 AM, "Sheng Yang" <sheng@yasker.org> wrote:
>>>
>>> >On Sun, Jan 13, 2013 at 11:21 PM, Hugo Trippaers <
>>> >HTrippaers@schubergphilis.com> wrote:
>>> >
>>> >> Side note, but where are we going to do development? I think we should
>>> >>get
>>> >> our work in a feature branch as soon as possible so we can all
>>> >>contribute
>>> >> to the development.
>>> >>
>>> >> Sheng, can you push any work you have done to a branch? I'll check and
>>> >>see
>>> >> if I need to commit any of my changes then.
>>> >>
>>> >
>>> >We'd like to have the work based on Chiradeep's network refactor branch,
>>> >but currently we're waiting for Javelin to be merged.
>>> >
>>> >And I am dong some PoC right now, so no code is written so far. I just
>>> >found it's not that straightforward to get the dnsmasq work as we
>>> > thought.
>>> >
>>> >We need DUID from client(not MAC) to hand out ipv6 addresses, but I am
>>> > not
>>> >sure if that's something we can know from mgmt server side.
>>> >
>>> >--Sheng
>>> >
>>> >>
>>> >> Cheers,
>>> >>
>>> >> Hugo
>>> >>
>>> >> -----Original Message-----
>>> >> From: John Kinsella [mailto:jlk@stratosec.co]
>>> >> Sent: vrijdag 11 januari 2013 19:50
>>> >> To: cloudstack-dev@incubator.apache.org
>>> >> Subject: Re: [ASFCS41] IPv6 Support
>>> >>
>>> >>
>>> >> On Jan 10, 2013, at 11:09 AM, Sheng Yang <sheng@yasker.org> wrote:
>>> >>
>>> >> > On Tue, Jan 8, 2013 at 10:57 PM, John Kinsella <jlk@stratosec.co>
>>> >>wrote:
>>> >> >> From a quick glance at it's man page, looks like it can do
v4 and v6
>>> >> >> leases at the same time...
>>> >> > You mean dhcpd?
>>> >> >
>>> >>
>>> >> No, dnsmasq looked like it could...
>>> >>
>>> >> > I am exploring all the possibility right now.
>>> >> >
>>> >> > Currently dnsmasq in our systemvm doesn't support DHCPv6, so we
would
>>> >> > either update to a newer version dnsmasq, or using other dhcp
>>> >>server(e.g.
>>> >> > dhcpd) on DHCPv6.
>>> >> >
>>> >> > But replacing the dhcp server is a big work, all the configuration
>>> >> > need to be rewritten. Regarding dhcpd, I haven't figured out how
much
>>> >> > effort we need to spend if we want to switch.
>>> >> >
>>> >> > There is one possible solution for this release: say, using dhcpd
>>> >> > only
>>> >> > for IPv6, to reduce our effort of introducing IPv6(if it's easier
>>> >> > than
>>> >> > moving to dhcpd). And then we can make the choice in the later
>>> >>release.
>>> >> >
>>> >> > John, do you have some experiences can share regarding dhcpd?
>>> >> >
>>> >> > Also, regarding your problem, have you used cloudstack to distribute
>>> >> > IP? I don't think we support leasing on two /28s in advance network
>>> >>now?
>>> >>
>>> >> So, in my lab env I've made a few changes to the server and UI to allow
>>> >> multiple IP blocks in basic network (haven't tried in advanced yet),
>>> >>then
>>> >> additionally I'm passing a netmask down to the agent, then that's
>>> >> passed
>>> >> through to dhcp_entry.sh, and then into edithosts.sh, which I've
>>> >>updated to
>>> >> support dhcpd.  The one thing I wanted to do but haven't is to update
>>> >> edithosts.sh to support both dnsmasq as well as dhcpd.
>>> >>
>>> >> Note: this was done in-house without community involvement as I didn't
>>> >> expect general interest in this, but I'm happy to use experience gained
>>> >>to
>>> >> write similar support in the ASF tree.
>>> >>
>>> >> Johh
>>> >>
>>>
>>

Mime
View raw message