Re: comment for draft-ietf-vrrp-unified-spec-00

"G. C." <[email protected]> Tue, 3 Jun 2008 21:08:24 -0400
Newsgroups gmane.ietf.vrrp
Message-ID <[email protected]>
Steve,

After reading the text around AdvSendAdvertisements in RFC 4861, the
impression I have is that turning this off may be confused by the reader to
indicate that **all** Router Advertisements will be suppressed for that
interface (including the RA proposed by draft 02).

Should we define a separate per-interface flag to indicate that Router
Advertisements using the interface MAC address as the link-layer option
would be suppressed or not?  Do we even need a configuration parameter for
this, i.e. for a master owner router, is there ever a need for both RAs?

Cheers,
Georges


On Tue, Jun 3, 2008 at 1:36 PM, Stephen Nadas <[email protected]>
wrote:

>  Hi Georges,
>
> this section says that advertising interface must send an RA if
> corresponding AdvSendAdvertisements flag is TRUE
>
> i agree that if this happens on a vrrp interface there may be confusion on
> host as to which RA to use. (too bad host cannot be smart enough to use vrrp
> mac)
>
> i think all that VRRP spec could possibly say is to note somewhere that
> this is possible and that when VRRP is in use, the corresponding
> AdvSendAdvertisements flag should be FALSE.
>
> i would like to hear other views from list, please
>
> Thanks,
> Steve
>
>  ------------------------------
> *From:* G. C. [mailto:[email protected]]
> *Sent:* Tuesday, June 03, 2008 1:25 PM
> *To:* Stephen Nadas
> *Cc:* [email protected]; [email protected]
>
> *Subject:* Re: [VRRP] RE: comment for draft-ietf-vrrp-unified-spec-00
>
> Hi Steve,
>
> Here's my understanding of the problem...
>
> RFC 4861 describes the concept of Advertising Interface (Section 6.2.2).
> As such an an advertising interface that belongs to a router would send out
> RAs with the interface's link-local address (LLA) as the source, and the
> interface MAC address as the source link-layer option.
>
> Consider the case where the interface owns addresses that would be
> advertised in a VRRP Advertisement.  This interface would then send out one
> RA as indicated in the previous paragraph and, based on draft 02, another
> one using the interface's LLA as the source address (since it's the address
> owner) and the virtual MAC address as the source link-layer option.
>
> Now, from the point of view of the host that receives these RAs:  RFC 4861
> Section 6.3.4 indicates that the host may update its Neighbor Cache with the
> link-layer address included in the RA.  Since the 2 RAs described above
> would indicate different link-layer addresses for the same IP address, the
> host's ND cache would be constantly be updated back and forth with these
> addresses.
>
> Regards,
> Georges
>
>
> On Tue, Jun 3, 2008 at 10:01 AM, Stephen Nadas <[email protected]>
> wrote:
>
>>  Georges, Tomohiko,
>>
>> I do not see in 2461 the words "master" or "backup" so it is not clear to
>> me what exactly is meant by Tomohiko's when he says (from below):
>>
>> > In other hand, according to RFC 2461, both Master and Backup
>> > routers are also able to send RA packets for themselves,
>> > whose source address is their interface IPv6 link-local
>> > address, and source link-layer address option is their
>> > interface MAC address.
>>
>> master and backup routers are VRRP terms - so I think I do not agree with
>> the above as I would expect VRRP routers to use VRRP Virtual MACs.
>>
>> Thanks,
>> Steve
>>
>>  ------------------------------
>> *From:* Georges Chung [mailto:[email protected]]
>> *Sent:* Monday, June 02, 2008 5:25 PM
>> *To:* Stephen Nadas
>> *Subject:* Re: [VRRP] RE: comment for draft-ietf-vrrp-unified-spec-00
>>
>>   Hi Steve,
>>
>> I now realize that I may have missed the boat on this one.
>>
>> My email's subject line reflected Tomohiko's original email subject line.
>> I'm afraid draft -02 still does not specify what to do in the situation
>> described by Tomohiko.
>>
>> Regards,
>> Georges
>>
>>
>> On Mon, Jun 2, 2008 at 11:40 AM, Stephen Nadas <
>> [email protected]> wrote:
>>
>>>  Hi Georges,
>>>
>>> I am a bit confused here - draft is at -02 and went thru wg last call.
>>> Is the current text in -02 satisfactory or not?
>>>
>>> Thanks,
>>> Steve
>>>
>>>  ------------------------------
>>> *From:* [email protected] [mailto:[email protected]] *On Behalf
>>> Of *Georges Chung
>>> *Sent:* Monday, June 02, 2008 9:33 AM
>>> *To:* [email protected]
>>> *Subject:* [VRRP] RE: comment for draft-ietf-vrrp-unified-spec-00
>>>
>>>   I think Tomohiko has a point.
>>>
>>> Sending a Router Advertisement with the interface's MAC as the source
>>> link-layer address would refresh the hosts' Neighbor Cache entry with the
>>> router's physical MAC address.  We want the hosts to use the virtual MAC
>>> address of the router instead.
>>>
>>> Can the draft be updated with this?
>>>
>>> Thanks,
>>> Georges
>>>
>>>
>>>
>>> ----------------------------------------------------------------------------------------
>>>
>>> Hello Tomohiko,
>>>
>>> Thank you for your comment.  I am copying to the vrrp list as well for
>>>
>>>
>>>
>>> any comments WG may have.
>>>
>>> Regards,
>>> Steve
>>>
>>> > -----Original Message-----
>>> > From: kura at iij.ad.jp [mailto:kura <kura> at iij.ad.jp]
>>>
>>>
>>>
>>> > Sent: Wednesday, January 23, 2008 12:54 AM
>>> > To: Stephen Nadas
>>> > Subject: comment for draft-ietf-vrrp-unified-spec-00
>>> >
>>> > Hello,
>>> >
>>> > This is Tomohiko Kurahashi from IIJ which is an ISP in Japan.
>>>
>>>
>>>
>>> > I have a comment for draft-ietf-vrrp-unified-spec-00, with
>>> > regard to the interaction with IPv6 Router Advertisements (RAs).
>>> >
>>> > According to the draft, Master router (only) sends RA packets
>>>
>>>
>>>
>>> > for a virtual router, whose source address is the IPv6
>>> > link-local address associated with the virtual router (e.g.
>>> > link-local address of address owner), and source link-layer
>>> > address option is the virtual router MAC address.
>>>
>>>
>>>
>>> >
>>> > In other hand, according to RFC 2461, both Master and Backup
>>> > routers are also able to send RA packets for themselves,
>>> > whose source address is their interface IPv6 link-local
>>> > address, and source link-layer address option is their
>>>
>>>
>>>
>>> > interface MAC address.
>>> >
>>> > Under such situation that both kinds of RAs mentioned above
>>> > are being advertised, some hosts receiving former (first) may
>>> > set their default route to the virtual router, and others
>>>
>>>
>>>
>>> > receiving latter (first) may set to one of real routers. This
>>> > must be an unwilling result for network managers/administrators.
>>> >
>>> > I think it should be mentioned explicitly in the draft that
>>>
>>>
>>>
>>> > VRRP routers MUST stop advertising the latter kind of RAs
>>> > automatically (or by configuration).
>>> >
>>> > FYI, the following is Cisco manual of HSRP for IPv6.
>>> > http://www.cisco.com/en/US/products/ps6350/products_configurat
>>>
>>>
>>>
>>> > ion_guide_chapter09186a008078f345.html#wp1066077
>>> >
>>> > I'm glad if you take my comment into consideration.
>>> >
>>> > Best regards,
>>> > --
>>> > Tomohiko Kurahashi <kura at iij.ad.jp>
>>>
>>>
>>>
>>> > Internet Initiative Japan Inc.
>>>
>>>
>>>
>>
>

_______________________________________________
vrrp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/vrrp