Re: [Softwires] [dhcwg] Is tunnelling DHCP multicast messages in 6rd unicast tunnel to BR acceptable for DHCP redundancy?//re: About draft-guo-softwire-6rd-ipv6-config-00

[Xu Xiaohu <[e-mail]>]

> -----邮件原件-----
> 发件人: Ted Lemon [mailto:[e-mail]]
> 发送时间: 2010年8月15日 1:15
> 收件人: Xu Xiaohu
> 抄送: 'Ole Troan'; [e-mail]; [e-mail]
> 主题: Re: [dhcwg] Is tunnelling DHCP multicast messages in 6rd unicast
tunnel
> to BR acceptable for DHCP redundancy?//re: [Softwires] About
> draft-guo-softwire-6rd-ipv6-config-00
> 
> On Aug 8, 2010, at 11:59 PM, Xu Xiaohu wrote:
> > In addition, in case that the DHCP
> > server/relay on a given BR is unavailable due to some reason (e.g., no
> > available address in the pool) but the other functions of it (e.g., the
> > anycast route for it) are still available, once the DHCP message is
tunneled
> > to that BR, the DHCP service is unavailable any more. That's to say,
it's
> > hard to achieve the high availability for DHCP servers/relay agents
since
> > the DHCP information-request can not be relayed to other DHCP
> > servers/relays.
> 
> I don't see the problem here.   If you want redundancy, you should
configure
> each BR to relay to more than one DHCP server.   This would be the default
anyway,
> according to RFC3315 section 20, which requires the relay to multicast if
not
> otherwise configured.

However, the BR itself (no matter as a relay agent or a DHCP server) is a
single point of failure, especially in the case where the DHCP service on
the BR is unavailable while the route for it (anycast address) is still OK. 

> > 2) The CPE as a DHCP relay agent sends the DHCP-relay-forward message to
one
> > of the learnt DHCP server (unicast) addresses. Since 6rd can support
unicast
> > flexible, the DHCP server can be located either on/behind any BR or
behind
> a
> > 6rd CPE which is owned by the 6rd SP.
> 
> This is completely contrary to the recommendations in RFC3315, which
require
> the client to multicast, and require any relays to multicast by default.
> Trying to relay DHCPv6 traffic over an IPv4 transport seems needlessly
> complicated.   You have the tunnel--why not use it?

Yes, the CPE could directly tunnel the DHCP request messages (multicast)
received from the clients to the nearest BR. However, since the source
addresses of these DHCP messages are link-local addresses, how could the BR
return the corresponding DHCP reply messages to the clients in the 6rd
scenario?

Best wishes,
Xiaohu