Hi Zhibo,
On 30/07/2020 15:18, Huzhibo wrote:
HI Peter：
1：No problem is a problem that never can be proved.
2：For bgp example，when the pe node down，the bgp peer must down within
30 mintus，It will not get it up via cancle advertise pua.
for the above it is sufficient to advertise the unreachability for few
seconds from each ABR independently. That would be a much more solid
proposal.
thanks,
Peter
ZHIBO
--------------------------------------------------
胡志波 Hu Zhibo
Mobile: +86-18618192287 <tel:+86-18618192287>
Email: huzh...@huawei.com <mailto:huzh...@huawei.com>
*发件人：*Peter Psenak <ppse...@cisco.com>
*收件人：*Huzhibo <huzh...@huawei.com>;Aijun Wang <wang...@chinatelecom.cn>
*抄 送：*Aijun Wang <wangai...@tsinghua.org.cn>;lsr
<lsr@ietf.org>;Xiaoyaqun <xiaoya...@huawei.com>
*时 间：*2020-07-30 21:02:49
*主 题：*Re: [Lsr] New Version Notification for
draft-wang-lsr-prefix-unreachable-annoucement-03.txt
Huzhibo,
On 30/07/2020 14:49, Huzhibo wrote:
Hi peter：
On 30/07/2020 14:28, Aijun Wang wrote:
Hi, Peter:
Aijun Wang
China Telecom
On Jul 30, 2020, at 20:00, Peter Psenak <ppse...@cisco.com> wrote:
Hi Aijun,
On 30/07/2020 13:44, Aijun Wang wrote:
Hi, Peter:
Currently, we have the following consideration:
1. If not all of the ABRs announce the unreachability of the specified
prefix(the specified prefix is still up), such PUA information will continue
advertising by the unreachable ABRs.
so the behavior of ABR is going to be dependent on the behavior of the the
other ABR(s)? That is a very thin ice you are dancing on.
[WAJ] it is easy for ABR to get these information and make the judgment. All
ABRs locate within the same area.
having that info is not the issue. The cyclic dependency is the one.
[HZB] The cyclic dependency does not occur. The PUA advertisement behavior of
an ABR is not affected by the transient status of other ABRs. The condition for
canceling PUA advertisement is that all ABRs advertise the PUA information with
a certain prefix for 30 minutes.
you have not convinced me with the above statement. You are making a
dependency on the ABR behavior based on the other ABRs' behavior. That's
a call for a trouble.
2. If all of the ABRs announce the PUA of the specified prefix, such
information will be announced for a configurable duration, to make sure other
protocol(for example BGP) that based on the specified prefix has converged.
well, the problem is that no mater what is your configured time to advertise
the un-reachability, you have no way of knowing whether the resource that
became unreachable have done so intentionally and whether it will ever come
back. And guessing based on the timer is not going to make it much better I'm
afraid.
[WAJ] We need not care bout the reason for the unreachability. When ABR get the
PUA information from also all other ABRs, it will start one timer to count the
duration of following PUA message. After the duration, all ABRs will stop the
advertisement.
once you stop advertising the un-reachability, you are back to square one as
you were with the summary only.
[HZB] There are two scenarios: First: The node is actually Down. In this case, upper-layer services will be converged 30 minutes earlier.
once you stop the unreachable advertisement, the resource will be seen
as reachable outside of its area, even when it is down. That is a
problem that you are trying to solve in a first place.
thanks,
Peter
Even if the PUA is canceled, the problem will not occur. Second: If the
node is not Down and only an ABR is unreachable, the ABR will
continuously advertises PUA.
thanks,
peter
thanks,
Peter
How about the above consideration and Do you have other thoughts ?
Aijun Wang
China Telecom
On Jul 30, 2020, at 17:21, Peter Psenak <ppsenak=40cisco....@dmarc.ietf.org>
wrote:
Hi Ajun,
one additional problem on top of others that have been mentioned is
how are you going to get rid of "old" un-reachability
announcements/
Let's imagine you have the following prefixes in your area 1:
- 10.10.0.1/32 - 10.0.0.255/32 - used for loopback adresses
- 10.10.1.0/30 - 10.10.10.252/32 - used for transit links
For simplicity your summary for area 1 is 10.0.0.0/16
1. Everything is up, you generate 10.0.0.0/16 on the ABR that
connects to area 1 to all remaining connected areas
2. A router in area 1 with loopback 10.10.0.25/32 goes down.
3. ABR in area 1 generates un-reachable announcement for
10.0.0.25/32 to all other connected areas
4. When does ABR stop generating the un-reachable announcement for 10.0.0.25/32? In section 6 you
mention that "The receiver router should keep the black hole routes for PUA as one
configurable time(MAX_T_PUA)", but the draft does not say anything about when the
un-reachability announcements should be removed by ABR. We can not rely on 10.10.0.25/32 ever
coming back, as the user may have simply removed it for good. Keeping the "stale"
un-reachability announcements may result in the LSDB growth over the period of time.
thanks,
Peter
On 27/07/2020 03:32, Aijun Wang wrote:
Hi, LSR experts:
We have uploaded the new version of this PUA(Prefix Unreachable Announcement)
draft. The main updates are the followings:
1) Describes the solution that using tunnel to redirect traffic among ABRs,
when all ABRs reaches the PUA limit.
2) Describe fast rerouting to avoid routing black hole.
3) Defining PUA capabilities announcements for OSPFv2/OSPFv3 and ISIS.
There are also some arguments about the current solution for PUA, for example:
1) Is it suitable to set the "Prefix Originator" sub-TLV to NULL to indicate
the prefix is unreachable?
2) if not, what's the consideration? What's the other convincible solution?
Wish to hear comments and suggestions on the above issues. We will also have
the presentation on the coming IETF LSR meeting.
Best Regards
Aijun Wang
China Telecom
-----Original Message-----
From: internet-dra...@ietf.org [mailto:internet-dra...@ietf.org]
Sent: Monday, July 27, 2020 9:16 AM
To: Zhibo Hu <huzh...@huawei.com>; Aijun Wang
<wang...@chinatelecom.cn>; Yaqun Xiao <xiaoya...@huawei.com>
Subject: New Version Notification for
draft-wang-lsr-prefix-unreachable-annoucement-03.txt
A new version of I-D,
draft-wang-lsr-prefix-unreachable-annoucement-03.txt
has been successfully submitted by Aijun Wang and posted to the IETF repository.
Name: draft-wang-lsr-prefix-unreachable-annoucement
Revision: 03
Title: Prefix Unreachable Announcement
Document date: 2020-07-27
Group: Individual Submission
Pages: 11
URL:
https://www.ietf.org/internet-drafts/draft-wang-lsr-prefix-unreachable-annoucement-03.txt
Status:
https://datatracker.ietf.org/doc/draft-wang-lsr-prefix-unreachable-annoucement/
Htmlized:
https://tools.ietf.org/html/draft-wang-lsr-prefix-unreachable-annoucement-03
Htmlized:
https://datatracker.ietf.org/doc/html/draft-wang-lsr-prefix-unreachable-annoucement
Diff:
https://www.ietf.org/rfcdiff?url2=draft-wang-lsr-prefix-unreachable-annoucement-03
Abstract:
This document describes the mechanism that can be used to announce
the unreachable prefixes for service fast convergence.
Please note that it may take a couple of minutes from the time of
submission until the htmlized version and diff are available at tools.ietf.org.
The IETF Secretariat
_______________________________________________
Lsr mailing list
Lsr@ietf.org
https://www.ietf.org/mailman/listinfo/lsr
_______________________________________________
Lsr mailing list
Lsr@ietf.org
https://www.ietf.org/mailman/listinfo/lsr
_______________________________________________
Lsr mailing list
Lsr@ietf.org
https://www.ietf.org/mailman/listinfo/lsr
