Re: [bess] EVPN MH: Backup node behavior in Primary Path Failure

["Rabadan, Jorge (Nokia - US/Mountain View)" <[e-mail]>]

Anush,

From: Anush Mohan <[e-mail]>
Date: Thursday, October 4, 2018 at 5:47 PM
To: "Rabadan, Jorge (Nokia - US/Mountain View)" <[e-mail]>
Cc: "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>
Subject: Re: [bess] EVPN MH: Backup node behavior in Primary Path Failure

Hi Jorge,

      Related to this topic, I had couple of queries as well. Could you please clarify.

1.  I hope the section of RFC pasted by Jai is superceded by the particular DF algorithm used. If all PEs can decide one particular backup PE for Ethernet-segment based on HRW (for e.g),
     only that particular backup-PE can be used for unicasting traffic. We can avoid flushing mac-entry in this case.
[JORGE] see my other email. In RFC7432 you can avoid mac flushing at the remote PEs only if there are two PEs in the ES, with more than two, the remote PEs need to flush the macs and flood:

   If there is only one backup PE for a given ES, the remote PE MAY use
   the primary PE's withdrawal of its set of Ethernet A-D per ES routes
   as a trigger to update its forwarding entries, for the associated MAC
   addresses, to point towards the backup PE.  As the backup PE starts
   learning the MAC addresses over its attached ES, it will start
   sending MAC/IP Advertisement routes while the failed PE withdraws its
   routes.  This mechanism minimizes the flooding of traffic during
   fail-over events.

   If there is more than one backup PE for a given ES, the remote PE
   MUST use the primary PE's withdrawal of its set of Ethernet A-D per
   ES routes as a trigger to start flooding traffic for the associated
   MAC addresses (as long as flooding of unknown unicast packets is
   administratively allowed), as it is not possible to select a single
   backup PE.

[JORGE] In RFC8214 there is a single backup PE, even with more than 2 PEs in the ES, and that condition is signaled. We’ve had some discussions to re-use this backup signaling in RFC7432 based EVIs, but it is not there in existing RFC7432 networks.


2.  If 'all-active' multihoming is used and a particular MAC is learnt from multiple PEs on an Ethernet-segment, should we use 'mac-ip' route label for load-balancing traffic or alias-label from
     'EAD/ESI' route. Or it doesn't matter.
[JORGE] IMO it doesn’t matter much if you use a label per MAC-VRF (or per-BD) on the ES PEs, since the labels will be the same anyway and at the egress you do a mac-lookup anyway…

Regards
Anush

On Thu, Oct 4, 2018 at 8:10 PM Rabadan, Jorge (Nokia - US/Mountain View) <[e-mail]<mailto:[e-mail]>> wrote:
Muthu,

About this:

Suppose we have a full mesh of iBGP sessions b/w the PEs, then who would withdraw the routes if the primary PE / DF itself fails? Instead, the BGP session would timeout causing the primary PE's neighbors to flush out the A-D routes received from the primary PE, right? This can take several seconds, isn't it?

No, not in the implementations I know of. Next Hop tracking will immediately detect that the PE is not in the network anymore and the routes will be invalidated. You can also bootstrap the BGP sessions to BFD.
But that has nothing to do with EVPN!.. it’s regular BGP.

Thx
Jorge

From: Muthu Arul Mozhi Perumal <[e-mail]<mailto:[e-mail]>>
Date: Thursday, October 4, 2018 at 1:14 PM
To: "Rabadan, Jorge (Nokia - US/Mountain View)" <[e-mail]<mailto:[e-mail]>>
Cc: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: Re: [bess] EVPN MH: Backup node behavior in Primary Path Failure

Please see inline..

On Thu, Oct 4, 2018 at 3:33 PM Rabadan, Jorge (Nokia - US/Mountain View) <[e-mail]<mailto:[e-mail]>> wrote:
In-line.

Thx
Jorge

From: Jaikumar Somasundaram <[e-mail]<mailto:[e-mail]>>
Date: Thursday, October 4, 2018 at 11:28 AM
To: "Rabadan, Jorge (Nokia - US/Mountain View)" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Cc: Jiang He <[e-mail]<mailto:[e-mail]>>, P Muthu Arul Mozhi <[e-mail]<mailto:[e-mail]>>
Subject: RE: [bess] EVPN MH: Backup node behavior in Primary Path Failure

Thanks Jorge for the quick reply.
Please find further question below.

From: Rabadan, Jorge (Nokia - US/Mountain View) <[e-mail]<mailto:[e-mail]>>
Sent: Thursday, October 4, 2018 1:52 PM
To: Jaikumar Somasundaram <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Cc: Jiang He <[e-mail]<mailto:[e-mail]>>; P Muthu Arul Mozhi <[e-mail]<mailto:[e-mail]>>
Subject: Re: [bess] EVPN MH: Backup node behavior in Primary Path Failure

Hi,


Questions:
1.       Will the node in backup mode forward the packet to CE?

[JORGE] as soon as it becomes DF it can forward packets to the CE. The backup node will have to run DF election upon the ES route withdrawal from the primary. If AC-DF is enabled, it can also react to the withdrawal of AD routes from the primary PE.

[Jai] Does that mean if any packet comes to a node that is still in the backup mode will get dropped, before the new DF election is complete? Why cant this be used as FRR? Or what is the use case of having backup node(s)?

[JORGE2] when the primary node fails, ES and AD routes are withdrawn.
Suppose we have a full mesh of iBGP sessions b/w the PEs, then who would withdraw the routes if the primary PE / DF itself fails? Instead, the BGP session would timeout causing the primary PE's neighbors to flush out the A-D routes received from the primary PE, right? This can take several seconds, isn't it?

Regards,
Muthu

The AD route withdrawal is an indication for remote nodes that they have to send traffic to the backup (for a given MAC) or to flush the MACs if there are more than 2 PEs in the ES. Around the same time or maybe earlier, the ES route withdrawal will make the backup PE take over as DF. So the overall convergence time will depend on how/when those two things happen in time. Only the DF PE can forward traffic. A non-DF can never forward traffic or there will be risk of duplicate packets.


2.       Will all the nodes in backup mode forward the packet before DF election?

[JORGE] Only the new DF can forward.



3. If they forward, how is duplicate packets handled, in this case?
[JORGE] see above.

My two cents..

Thanks.
Jorge



From: BESS <[e-mail]<mailto:[e-mail]>> on behalf of Jaikumar Somasundaram <[e-mail]<mailto:[e-mail]>>
Date: Thursday, October 4, 2018 at 10:03 AM
To: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Cc: Jiang He <[e-mail]<mailto:[e-mail]>>, P Muthu Arul Mozhi <[e-mail]<mailto:[e-mail]>>
Subject: [bess] EVPN MH: Backup node behavior in Primary Path Failure


Hello Everyone,

Sorry if it is a duplicate. I repost this query as I did not receive any response yet.
(I was wondering if this mail already reached the group or not)


I have a question on Primary PE encountering a failure in EVPN multihoming

in single active mode.



RFC7432, section 14.1.1:

<snip>

   If there is more than one backup PE for a given ES, the remote PE

   MUST use the primary PE's withdrawal of its set of Ethernet A-D per

   ES routes as a trigger to start flooding traffic for the associated

   MAC addresses (as long as flooding of unknown unicast packets is

   administratively allowed), as it is not possible to select a single

   backup PE.

</snip>



Questions:

1. Will the node in backup mode forward the packet to CE?

2. Will all the nodes in backup mode forward the packet before DF election?

3. If they forward, how is duplicate packets handled, in this case?



Please help me anwere these questions.



Thanks & Regards

Jaikumar S


_______________________________________________
BESS mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/bess
_______________________________________________
BESS mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/bess