Re: [bess] Query/comments on draft-ietf-bess-evpn-lsp-ping-05

["Dikshit, Saumya" <[e-mail]>]

Hi Parag,

Thank you for the response. Please see inline with tag [SD2] and provide your further inputs.

Thanks
Saumya.

From: Parag Jain (paragj) [mailto:[e-mail]]
Sent: Saturday, September 11, 2021 8:19 PM
To: Dikshit, Saumya <[e-mail]>om>; [e-mail]; [e-mail]
Cc: [e-mail]
Subject: Re: Query/comments on draft-ietf-bess-evpn-lsp-ping-05

Hi Saumya

Pls see inline.

From: "Dikshit, Saumya" <[e-mail]<mailto:[e-mail]>>
Date: Thursday, September 9, 2021 at 3:54 AM
To: "Parag Jain (paragj)" <[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]>>
Cc: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: RE: Query/comments on draft-ietf-bess-evpn-lsp-ping-05

Hi Parag,

Please see inline. Let me know your thoughts.

Thanks
Saumya.

From: Parag Jain (paragj) [mailto:[e-mail]]
Sent: Wednesday, September 8, 2021 11:43 PM
To: Dikshit, Saumya <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Cc: [e-mail]<mailto:[e-mail]>
Subject: Re: Query/comments on draft-ietf-bess-evpn-lsp-ping-05

Hi Saumya

Pls see inline.

From: "Dikshit, Saumya" <[e-mail]<mailto:[e-mail]>>
Date: Tuesday, September 7, 2021 at 3:22 PM
To: "Parag Jain (paragj)" <[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]>>
Cc: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: RE: Query/comments on draft-ietf-bess-evpn-lsp-ping-05

Hi Parag,

Thanks for the response. I have few bullets on the same.
Please help clarify and if there is a need to call them out explicitly.


  1.  “Consistency checkers” feature-set does validates the CP-DP parity and can be leveraged via management interface to the box.
     *   Do you imply the consistency check between protocol RIB and the dataplane FIB, Or the consistency between Software FIB (slow path) and the LC-FIB
Paragj> CP would mean BGP/EVPN/RIB which ever CP component has the info included in the sub-TLVs.
[SD] I am little unclear, as to how running the Sub-TLV parameters through the RIB, will ensure that this RIB entry (NLRI) was CHOSEN as the FIB entry.
Essentially, the RIB entry mapping to the Sub-TLV, has to contend with other RIB entries and also with same route published by other protocols (or instances of protocol), eventually get picked as FIB entry.  Lsp ping to the sub-tlv may pan out differently in RIB and in FIB. But as I understand, that is not the purpose of this reachability check defined in this draft.

Paragj2> I also mentioned bgp and evpn CP components above.

We should call out this out specifically in the document or stick to validating the datapath.
Paragj2> DP-CP consistency check is an important part of lsp ping functionality. As RFC8029 states, the LSP echo message contains sufficient information to  verify correctness of DP operations and verify DP against CP to localize the fault.
[SD2] I am not contending DP-CP validation when needed, but when partial information is known (w.r.t), it will be good to go with remaining parameters as wild/card. Even RFC8029 provides some leeway in various sections. For example, in https://datatracker.ietf.org/doc/html/rfc8029#section-4, it tends to keep the underlay information open-ended if not known:

“If the underlying (LDP) tunnel were not known, or
   was considered irrelevant, the FEC Stack could be a single element
   with just the VPN IPv4 sub-TLV.”


Both RIB and FIB validation (leading to successful echo response), may not indicate that right RIB and FIB are consistent with respect to sub-tlv being carried.
I suggest that we keep lsp ping to just the data plane validation. Consistency-check will require lot of compute (might need a complete path calculation of BGP-EVPN), to indicate the consistency. Good to know if there are reference implementations optimizing this already.
This is one more reason to use wild card.


  1.  Parameters such as RD, shall not make it to the DP and their presence is restricted to the NLRI (entries/tables) in the protocol RIB.
     *   In case the RIB specific parameters need validation, then on receive side processing of ping, should run it through the RIB and FIB both ?
Paragj> yes.

     *   In case it’s just the dataplane validation (which I can gather from this draft), then RIB validation is not required and RD’s  can carry “don’t care”.
  1.  If a need be, to perform “reachability-check to a tenant vrf (EVI) on remote NVE”, for which no route has been published yet ?
Paragj> only vrf-existence is not checked by lsp ping.
[SD] That’s a good solution to have. I have mentioned the use-case in below email.
I propose that we leverage the existing “EVPN IP Prefix Sub-TLV”, with appropriate values (may be wild-card/don’t care) to realize this.

Paragj2> EVPN IP Prefix sub-tlv is for verifying ip prefix in a vrf. I am not sure it should/can be applied to the use case you  mention.
[SD2] My take was to re-use tlvs/info carried in lsp-ping as already defined in this draft. If not agreeable to authors and group members, it will be good to define a new tlv (or otherwise) via an ancillary draft if needed. I can do that if, authors of this draft feel that it’s a misfit in this document. Since the label encoding can implicitly map to the VRF/EVI on the target, a sub-tlv indicating an EVI-check (either L2 or L3) can help the cause..

Thanks
Parag


as I mentioned in #2 of below email

     *   Is it possible to achieve that with lsp-ping check with existing sub-TLVs without “wild-card/don’t-care”

Thanks
Saumya.

From: Parag Jain (paragj) [mailto:[e-mail]]
Sent: Tuesday, September 7, 2021 11:56 PM
To: Dikshit, Saumya <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Cc: [e-mail]<mailto:[e-mail]>
Subject: Re: Query/comments on draft-ietf-bess-evpn-lsp-ping-05

Hi Saumya

The remote PE router processing the lsp ping packet, does consistency checks between data plane and control plane. RD, ESI fields along with other fields defined in the sub-tlvs are used for that purpose. Wildcard/don’t care values for these fields will defeat the purpose of DP-CP consistency checks.

Thanks
Parag

From: "Dikshit, Saumya" <[e-mail]<mailto:[e-mail]>>
Date: Thursday, September 2, 2021 at 1:42 PM
To: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Cc: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: Query/comments on draft-ietf-bess-evpn-lsp-ping-05
Resent-From: <[e-mail]<mailto:[e-mail]>>
Resent-To: <[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]>>
Resent-Date: Thursday, September 2, 2021 at 1:42 PM

[sending the queries in a different email with changed subject line]

Hello Authors of draft-ietf-bess-evpn-lsp-ping draft,

I have following queries regarding this draft:

>>>> Do we intend-to-use/call-out-usage-of “wild-card/don’t-care” values for attributes carried in the sub-TLVs ?
For example, If the admin intends  to check the reachability to host (MAC_X/IP_X) published (in route-type-2)  by remote PE.
The remote PE learnt it locally over ESI_X against Vlan X (mapped to BD_XYZ).

Is it possible, that the “EVPN MAC sub-tlv”  can carry the “Route Distinguisher” and “Ethernet Segment Identifier” as don’t care.

>>>> Another caseto handle would be test the reachability to tenant-VRF VRF_X (with EVPN mapped EVI) configured on the remote PE, PE1.
VRF_X has no active IP/IPv6 interface configured and its sole usage is to obtain the leaked (via IVRL) routes from other VRFs (non-EVPN) and PE1 published this to other peers via EVPN control plane. Till the first prefix (learnt ) route is published (Route Type 5) by PE1 for the EVI (mapped to VRF_X), the tunnels will not be provisioned on other PEs.
In order to test the reachability to VRF_X (on PE1) from another PEs, let’s say, PE2 or a centralized-controller (which can emulate/supports MPLS),

It may need to carry all/subset-of attributes with “don’t-care/wild-card” in “EVPN IP Prefix Sub-TLV”.


Please let know your thoughts on above.

Thanks
Saumya.