BGP connection flapping - Optional Attribute Error

[Evgen_Ch]

Hello.

Many days trying to solve the problem. Help me please. 

Route Reflector - Cisco 3945 (ip 172.22.128.128)

Client - Huawei S5710-EI (ip 172.22.182.190)

I need to connect my S5710-EI to Route Reflector (Cisco) via MP-BGP. Connection state up in  "Estabished" but immediatly turn down to state "Idle".

P.S. BGP configuration in add.

1) Logs massege on S5710-EI tell me:

Oct 18 2018 22:19:51+05:00 US-R2 %BGP/3/STATE_CHG_UPDOWN(l)[60]:The status of the peer 172ublic, StateChangeReason=VPNv4 NLRI Parsed Error)

Complete code error - Update Messege Error (code 3), Optional attribute error (subcode 9).

2) BGP errors on S5710-EI:

S5710-EI>display bgp errors discard

27053. 2018-10-31 23:54:16+05:00 Neighbor: 172.22.128.128

Reason : Invalid RD value received.
Error data : 28 00
27054. 2018-10-31 23:54:16+05:00 Neighbor: 172.22.128.128
Reason : The VPNv4 prefix length is error.

3) Logs massege on Cisco 3945 tell me:

Oct 18 09:08:00.150 UFA: %BGP-5-ADJCHANGE: neighbor 172.22.182.191 Down BGP Notification received

Analisis tell me, that RR Cisco after BGP connection to client turn to "Establish" send some Update massege. This UPDATE message contain incorrect optional attribute and huawei S5710 is send NOTIFICATION with error code 3 subcode 9. 

I think that some vpnv4 prefix in some UPDATE from RR Cisco has an incorrect format. But how its fix? All RD are normal and options - "send-community extended" activate on RR.

[Georg Pauwen]

Hello,

looking at your configs, the Huawei is not sending extended communities. Change:

peer iBGP_cluster advertise-community

to

peer iBGP_cluster advertise-ext-community

[Evgen_Ch]

Thank you for answer, but we tried and nothing was changed. We activate this command on both - RR and client. BGP session still flapping.

[Georg Pauwen]

Hello,

can you ping 172.22.128.128 from the Huawei ? Post the full configuration of the Huawei...

[paul driver]

Hello

On the RRC are you sure you are not inadvertently peering with another RRC??

As an interim you could negate such peering effecting the bgp update between RRC

RRC
router bgp 64552
no bgp client-to-client reflection

Also can you post the following:
sh ip route ospf | in 172.22.182.191

sh bgp neighbor 172.22.182.191

[Evgen_Ch]

Hello,

Peer 172.22.182.191 its another one peer (backup router for 172.22.182.190, S5710-EI too). Configuration are the same (exept IP). Connection RR to 172.22.182.191 is flapping from Establish-to-Idle-to-Establish too.

[paul driver]

Hello

So just to confirm you have two routers
172.22.182.190 & 172.22.182.191 peering with each other (ibgp) and each ibgp peering with the RR ?

[Evgen_Ch]

We have two routers:
172.22.182.190 ibgp peer RR (route-reflector-client)
172.22.182.191 ibgp peer RR. (route-reflector-client)

.190 and .191 not ibgp neighbors. Configurations of bgp sessions are the same. BGP flapping on the both routers.

[paul driver]

Hello

Have you checked you NLRI between the RR and the RRC>
Make sure you have a stable connection to/from these clients each ibgp peering ip from within ospf?

[Georg Pauwen]

Hello,

are IP addresses 172.22.182.191 and 172.22.182.190 loopback addresses on the Huawei ? If so, try and configure the 'peer connect-interface' command on the Huawei, specifiying the loopback...

[Evgen_Ch]

Hello,

My mistake - i wrote in the first post that client ip address 172.22.182.190, but I aded a configuration BGP for 172.22.182.191. Excuse me.

But in really - it is similar routers that reserve each other. BGP configurations and problems are the same.

In add - full configuration for client 172.22.182.190. And as we can see in BGP section:

peer iBGP_cluster connect-interface LoopBack0
peer 172.22.128.128 as-number 64552
peer 172.22.128.128 group iBGP_cluster

On Cisco RR are the same:

neighbor iBGP_RR update-source Loopback1

neighbor 172.22.182.190 peer-group iBGP_cluster

[Georg Pauwen]

Hello,

what if you add the line in bold:

ipv4-family unicast
undo synchronization
import-route direct
peer 172.22.128.129 enable
peer 172.22.128.150 enable
peer iBGP_cluster enable
peer iBGP_cluster route-update-interval 1
peer 172.22.128.128 enable

peer 172.22.128.128 next-hop-local
peer 172.22.128.128 group iBGP_cluster

[Georg Pauwen]

Hello,

on the Cisco, remove:

neighbor iBGP_cluster update-source Loopback1

and add:

neighbor iBGP_cluster next-hop-self

On the Huawei, remove:

peer iBGP_cluster connect-interface LoopBack0

and add:

peer iBGP_cluster next-hop-local

[paul driver]

Hello

---

@Georg Pauwen wrote:

on the Cisco, remove:

neighbor iBGP_cluster update-source Loopback1 <------Georg  if this is removed then the peer wont even establish as bgp will think the peer  (loopback interface) is directly connected , So its requirement on both sides of a bgp peering when the neighbor peer ip isn't directly adjacent 

---

[Georg Pauwen]

The idea is to peer with the physical links and not the loopbacks. I have a feeling there might be a compatibility issue between Huawei and Cisco, because on paper the config looks good...