Bruno,
please see inline (##PP2):
On 26/07/2023 23:34, bruno.decra...@orange.com wrote:
Peter,
From: Peter Psenak <ppse...@cisco.com>
Sent: Wednesday, July 26, 2023 11:26 PM
Bruno,
please see inline (##PP):
On 26/07/2023 22:46, bruno.decra...@orange.com wrote:
Peter,
Please see inline
From: Peter Psenak <ppse...@cisco.com>
Bruno,
please see inline:
On 25/07/2023 21:11, bruno.decra...@orange.com wrote:
Peter,
Please see inline
From: Peter Psenak <ppse...@cisco.com>
Sent: Tuesday, July 25, 2023 6:49 PM
Bruno,
please see inline:
On 25/07/2023 18:36, bruno.decra...@orange.com wrote:
Peter,
Thanks for your answer.
Please see inline [Bruno]
From: Peter Psenak <ppse...@cisco.com>
Sent: Tuesday, July 25, 2023 6:05 PM
Bruno,
On 25/07/2023 14:39, bruno.decra...@orange.com wrote:
Hi all,
With RC5305, in IS-IS we can advertise two states for a prefix IP1:
* Positive reachability (state “1”), by advertising IP1 in TLV 135
with a metric lower than 0xFE000000
* No reachability (state “0”) by either:
o Not advertising IP1 in TLV 135
o Advertising IP1 in TLV 135with a metric larger than 0xFE000000
right, one being implicit the other one explicit.
Here you agree that advertising IP1 in TLV 135with a metric larger than
0xFE000000 is equivalent to No reachability / state “0” / Not advertising IP1
in TLV 135
For unreachable prefix announcement, we need to advertise a new state:
negative reachability (state “-1”) because advertising no reachability
(state “0”) is not enough since there is a covering prefix advertised.
Hence we need a new signaling.
draft-ppsenak-lsr-igp-ureach-prefix-announce is proposing to advertise
IP1 unreachability by advertising IP in TLV 135 with a metric larger
than 0xFE000000.
This does not work as this signal is already defined to advertise state
“0” (no reachability) while we want to advertise negative reachability
(state “-1”): one can’t signal two different states with a single signal.
This seems easy to fix in the last version of the draft: one just have
to signal state “-1” with the flags already defined and advertised.
In short, negative reachability would be signal by advertising IP1 in
TLV 135with a metric larger than 0xFE000000 and flag U (Unreachable
Prefix) or UP (Unreachable Planned Prefix) set.
that is exactly what the draft describes. I'm not sure what is your point.
On the receiving side, state “-1” is learned from the reception of
either flag.
This is a very simple change, and it would address my concern.
Can this be done?
I'm not sure what exactly are you asking for as the draft is already
describing what you are asking for.
[Bruno] Good if this is what is meant in the draft and if this is only
misunderstanding.
To be specific, I'm asking about the following changes for IS-IS (please feel
free to reformulate, OSPF is up to you)
§5.4
OLD:
U-Flag and UP-Flag are optional flags that have informative character.
Treatment of these flags on the receiver is optional and the usage of them is
outside of scope of this document.
NEW:
The setting of the U-Flag or the UP-Flag signals that the prefix is
unreachable. They constitute the UPA signals.
I don't think above is necessary:
[RFC5305] defines the encoding for advertising IPv4 prefixes using 4
octets of metric information. Section 4 specifies:
"If a prefix is advertised with a metric larger then MAX_PATH_METRIC
(0xFE000000, see paragraph 3.0), this prefix MUST NOT be considered
during the normal SPF computation. This allows advertisement of a prefix
for purposes other than building the normal IP routing table. "
Agreed with the quote.
"prefix NOT considered during normal SPF computation" does not change/affect
the outcome of the SPF computation. Hence does not remove reachability if a covering
prefix is present.
right. And UPA treatment is no different. It is only used to signal
unreachability to apps that are interested. It does not change anything
in forwarding itself.
UPA changes/stop forwarding to BGP prefix using the UPA prefix as netxt-hop.
This is a change in forwarding.
##PP
If the egress PE goes down BGP would move away to alternate NH - all we
do with UPA is to let BGP know sooner. Processing of the UPA is optional
and you are free ignore it and wait till the BGP finds out the remote PE
is dead. I see no backward compatibility problem here.
Plus at the beginning of this email, you agreed that this signaling translate into state
"0". (absence of reachability signal)
I don't see how you can now say that this signaling translate in state "-1"
(negative reachability)
As a result of the above, from the IP routing perspective, prefix with a
metric larger then MAX_PATH_METRIC (0xFE00000) MUST be treated as
unreachable.
a) This prefix is absent in the RIB/FIB (just like if it were not advertised in
TLV)
b) Covering prefix is in the RIB/FIB.
c) Net result is (positive) reachability.
Could you point to the line(s) that you disagree with?
I agree, and exactly same happens with UPA in terms of forwarding.
Let's have IP2 being a BGP prefix having IP1 as BGP NH
Today, if IP1 is advertised with a metric larger then MAX_PATH_METRIC
IP2 is forwarding.to IP1
With UPA, if IP1 is advertised with a metric larger then MAX_PATH_METRIC
IP2 stops forwarding to IP1
So the exact same signaling is triggering different/opposite behavior
I would call this a change in the spec. And not backward compatible
##PP
please see my above response.
If the implementation does not do that, it is clearly
violating RFC5305.
We added the new flags, to distinguish from other cases where the prefix
might be advertised with metric higher than 0xFE000000, whatever that
use case is. But the meaning of the metric higher than 0xFE000000 from
the IP routing perspective does not changes by the new flags.
This is exactly the concerned that I raised in my original email. (the one you
did not see the difference between my point and the draft).
I'm still not clear what is your concern.
Let's have IP2 being a BGP prefix having IP1 (PE1) as BGP NH
Today, in order to trigger a periodic or on-demand monitoring of this PE (e.g.
BFD, ping, sr-mpls monitoring (rfc8403)..) I periodically advertises IP1 with a
metric higher than 0xFE000000. This triggers on a centralized (or distributed
system) the monitoring toward PE1.
I'm free to do this and this is by the spec.
With UPA, this would suddenly trigger unreachability signaling to this PE,
breaking IP2 forwarding to PE1.
##PP
so you are saying that you can use metric higher than 0xFE000000 for
your monitoring case, but not for UPA?
##Bruno
My usage does not affect BGP routing. UPA does.
##PP2
so what's the point of monitoring something when you take no action if
the device dies?
Anyway, UPA processing is optional, so you have a choice to ignore it.
Similar to running a BFD session to the remote PE - you can take action
when the BFD goes down, or you can ignore it.
Any usage of such metric was
non-standard so far.
##Bruno
I would argue that the use and consequences of metric higher than 0xFE000000 is
well specified in RFC 5305.
##PP
RFC 5305 says "allows advertisement of a prefix for purposes other than
building the normal IP routing table". All UPA does is that it actually
defines one such "purpose" (and the first one, BTW). Anything before
that was private and non-standard.
For UPA we defined flags to signal the UPA case for such metric usage,
which distinguishes it from your "unspecified" case. As a result UPA
does not conflict with your private use of such metric.
##Bruno
" U-Flag and UP-Flag are optional flags that have informative character."
They can't be relied upon to distinguish the UPA case.
I'm calling this a non-backward compatible change.
##PP
I'm not sure I see the backward compatibility issue here, given that UPA
can be distinguish from any other private use case of the metric higher
than 0xFE000000.
##Bruno
How can the receiver distinguish UPA signaling from other uses of metric higher
than 0xFE000000?
(flags are not reliable, see above).
All I'm asking is to make such Flags mandatory, and the UPA signal. (metric
0xFE000000 being only the transporter)
##PP
If the UPA is the only reason why such metric would be used with prefix
advertisement in your network (which is the case for 99,9% of the
networks) the flags would not be needed. That's why we keep them optional.
thanks,
Peter
Thanks,
--Bruno
thanks,
Peter
Thanks,
--Bruno
thanks,
Peter
Thanks,
--Bruno
§5
"Even though in all cases the treatment of such metric, or NU-bit, is specified for
IS-IS, OSPF and OSPFv3, having an explicit way to signal that the prefix was advertised
in order to signal unreachability is useful to distinguish it from other cases where the
prefix with such metric is advertised."
:s/useful/required
§Abstract
OLD: This document describes how to use existing protocol mechanisms in IS-IS
and OSPF to advertise such prefix reachability loss.
NEW: This document defines two new flags in IS-IS and OSPF to advertise such
prefix reachability loss. In order to support incremental deployment, such
advertisement use a high metric value already having special meaning of OSPF
and IS-IS.
§1
OLD: This document describes how the use of existing protocol mechanisms can
support the necessary functionality without the need for any protocol
extensions. The functionality being described is called Unreachable Prefix
Announcement (UPA).
NEW: This document defines two new flags in IS-IS and OSPF to the necessary
functionality without the need for any protocol extensions. The functionality
being described is called Unreachable Prefix Announcement (UPA).
please see my above response. The treatment of the metric higher that
0xFE000000 and LSInfiity has been defined long time back and is not
affected by the newly defined flags.
thanks,
Peter
Thank you,
Regards,
--Bruno
The drawback is that the existing pre-standard implementation would need
to be updated. But that’s the rule for a pre-standard implementation.
not really, the handling of the metric higher that 0xFE000000 is already
defined in the existing spec.
regards,
Peter
The faster we do the change, the easier it is, especially for other
implementations.
--Bruno
Orange Restricted
____________________________________________________________________________________________________________
Ce message et ses pieces jointes peuvent contenir des informations
confidentielles ou privilegiees et ne doivent donc
pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce
message par erreur, veuillez le signaler
a l'expediteur et le detruire ainsi que les pieces jointes. Les messages
electroniques etant susceptibles d'alteration,
Orange decline toute responsabilite si ce message a ete altere, deforme ou
falsifie. Merci.
This message and its attachments may contain confidential or privileged
information that may be protected by law;
they should not be distributed, used or copied without authorisation.
If you have received this email in error, please notify the sender and delete
this message and its attachments.
As emails may be altered, Orange is not liable for messages that have been
modified, changed or falsified.
Thank you.
_______________________________________________
Lsr mailing list
Lsr@ietf.org
https://www.ietf.org/mailman/listinfo/lsr
