Hi, Gyan:

I am not sure what you meant by "non-deterministic" behavior. If you
referred to dampening the IGP metric export, there is this knob "bgp
dynamic-med-interval".

If one wants symmetric routing with scale at the prefix-level, "exporting
IGP metric as MED" is the best option.

Thanks.   -- Enke


On Mon, Oct 18, 2021 at 1:50 PM Gyan Mishra <[email protected]> wrote:

>
> Hi Enke
>
> Understood.  Sending IGP metric into BGP can have unwanted consequences as
> MED is higher up in BGP path selection and so can cause much non
> deterministic behavior.  It is better to not match internal but rather do
> set metric to set higher MED on all prefix to make the path preferred.  Or
> you can de-prefer the path by setting metric/med to 0.  In my experience
> there are better ways such as prepending and local preference to maintain
> symmetry then exporting IGP metric to BGP.
>
> Kind Regards
>
> Gyan
> On Mon, Oct 18, 2021 at 1:50 PM Enke Chen <[email protected]>
> wrote:
>
>> Hi, Gyan:
>>
>> > IGP metric exported to BGP as MED does not come into play here as we
>> are not redistribution ISIS into BGP.
>> > We are doing next hop self on egress PE and it’s the PE loopback next
>> hop attribute that is flooded.
>>
>> "Exporting IGP metric to BGP" (also discussed in Sect. 1.1, RFC 5302) is
>> different from "redistributing IGP into BGP". In Cisco IOS the feature is
>> configured as "set metric-type internal" in an outbound route-map.
>>
>> Thanks.   -- Enke
>>
>> On Sat, Oct 16, 2021 at 9:35 AM Gyan Mishra <[email protected]>
>> wrote:
>>
>>>
>>> Hi Enke
>>>
>>> See section 5 of PUA draft use case.
>>>
>>>
>>> https://datatracker.ietf.org/doc/html/draft-wang-lsr-prefix-unreachable-annoucement-07#section-5
>>> <https://urldefense.proofpoint.com/v2/url?u=https-3A__datatracker.ietf.org_doc_html_draft-2Dwang-2Dlsr-2Dprefix-2Dunreachable-2Dannoucement-2D07-23section-2D5&d=DwMFaQ&c=V9IgWpI5PvzTw83UyHGVSoW3Uc1MFWe5J8PTfkrzVSo&r=OPLTTSu-451-QhDoSINhI2xYdwiMmfF5A2l8luvN11E&m=EP67fgSmvoUQ60_OUUqc0BJ3zy3bmRVf8DrsucnyeGk&s=juwHypuac_GTZeZhN_M7ms5blX0_6WSIt7YGOCKzJe4&e=>
>>>
>>> Also see RFC 5283 LDP Inter-area extension and use of LPM matching
>>> summarization and issue with black hole routing with RFC 5283 summarization
>>> gap that exists when the BGP next hop attribute LPM prefixes go away do fo
>>> PE node failover.
>>>
>>> Also RFC 5302 ramifications of domain wide flooding of loopbacks
>>> scalability issue for large areas and solution needed to make RFC 5283 LPM
>>> summarization possible.
>>>
>>> Operators don’t want to flood the loopbacks domain wide for scalability
>>> reasons.
>>>
>>> This issue with LPM matching as well for IP or SRv6 based networks when
>>> component prefix goes away from node failure when summarization is utilized.
>>>
>>> The issue with multiple multiple paths can be resolved with ECMP path
>>> between areas with multiple level1-2 routes redundant paths so that you
>>> don’t run into symmetry issues.  Also their are cases where symmetry is not
>>> necessary.  Also you can use hardcoded metrics to set preferred path.  IGP
>>> metric exported to BGP as MED does not come into play here as we are not
>>> redistribution ISIS into BGP.  We are doing next hop self on egress PE and
>>> it’s the PE loopback next hop attribute that is flooded.
>>>
>>> Kind Regards
>>>
>>> Gyan
>>>
>>> On Fri, Oct 15, 2021 at 1:21 PM Enke Chen <[email protected]>
>>> wrote:
>>>
>>>> Hi, Folks:
>>>>
>>>> As I recall, the loopbacks and their metrics need to be propagated
>>>> individually to the whole network (including leaking from Level 2 to Level
>>>> 1) for the following reasons:
>>>>
>>>> 1) Path selection: in case there are multiple Level 2 /  Level 1
>>>> routers in an area.
>>>> 2) Export IGP metric : when the IGP metrics are exported to BGP as MED
>>>> for symmetric routing.
>>>>
>>>> Not sure if there is still a problem to be solved.
>>>>
>>>> Regards,
>>>>
>>>> -- Enke
>>>>
>>>> ---------- Forwarded message ----------
>>>> From: Tony Li <[email protected]>
>>>> To: "Les Ginsberg (ginsberg)" <[email protected]>
>>>> Cc: Gyan Mishra <[email protected]>, lsr <[email protected]>
>>>> Bcc:
>>>> Date: Wed, 13 Oct 2021 22:50:33 -0700
>>>> Subject: Re: [Lsr] "Prefix Unreachable Announcement" and "IS-IS and
>>>> OSPF Extension for Event Notification"
>>>>
>>>> Les,
>>>>
>>>> If you’re advertising loopbacks that are outside of the summarized
>>>> space, then you end up with reachability and liveness.  Yes, there’s a cost
>>>> in scalability… it ain’t free.
>>>>
>>>> Tony
>>>> _______________________________________________
>>>> Lsr mailing list
>>>> [email protected]
>>>> https://www.ietf.org/mailman/listinfo/lsr
>>>> <https://urldefense.proofpoint.com/v2/url?u=https-3A__www.ietf.org_mailman_listinfo_lsr&d=DwMFaQ&c=V9IgWpI5PvzTw83UyHGVSoW3Uc1MFWe5J8PTfkrzVSo&r=OPLTTSu-451-QhDoSINhI2xYdwiMmfF5A2l8luvN11E&m=EP67fgSmvoUQ60_OUUqc0BJ3zy3bmRVf8DrsucnyeGk&s=g9Thil3w---hguNz-FkPmxjLHnurnmrUFzDHaSm1jaI&e=>
>>>>
>>> --
>>>
>>>
>>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__www.verizon.com_&d=DwMFaQ&c=V9IgWpI5PvzTw83UyHGVSoW3Uc1MFWe5J8PTfkrzVSo&r=OPLTTSu-451-QhDoSINhI2xYdwiMmfF5A2l8luvN11E&m=EP67fgSmvoUQ60_OUUqc0BJ3zy3bmRVf8DrsucnyeGk&s=IiPSPRNx8poa0PgzIswRH-Tmd8aPbgntkDetpL9kPdc&e=>
>>>
>>> *Gyan Mishra*
>>>
>>> *Network Solutions A**rchitect *
>>>
>>> *Email [email protected] <[email protected]>*
>>>
>>>
>>>
>>> *M 301 502-1347*
>>>
>>> --
>
>
> <https://urldefense.proofpoint.com/v2/url?u=http-3A__www.verizon.com_&d=DwMFaQ&c=V9IgWpI5PvzTw83UyHGVSoW3Uc1MFWe5J8PTfkrzVSo&r=OPLTTSu-451-QhDoSINhI2xYdwiMmfF5A2l8luvN11E&m=gtx8sc-_Jh00dX84ZuQEFj2HYti4M3KJRg7fxox4MpQ&s=omEgWRPZkDBZvAYi_C_wSKbG5YiBS-7XK703pp6uiM8&e=>
>
> *Gyan Mishra*
>
> *Network Solutions A**rchitect *
>
> *Email [email protected] <[email protected]>*
>
>
>
> *M 301 502-1347*
>
>
_______________________________________________
Lsr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lsr

Reply via email to