cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1694
Views
0
Helpful
14
Replies

OSPF Area 0 adjacency stay Extart or exchange

AllezLom63194
Level 1
Level 1

Hello the Community,

 

You can find my network topology on the attached file.

 

All the 5K are:

- Dell L3 switch

- OSPF area 0 configured

- each L3 switch has one IP adresse on one interface VLAN910

- each L3 switch has one "network 10.236.0.0/24 area 0" configured under router ospf 1

 

The VLAN910 carry only one IP network (10.236.0.0/24),

The VLAN910 is configured on each physical port and port-channel interfaces between all the 5K

"VLT domain" is a DELL feature that make on virtual switch composed with 2 physical switch

The switch 5K_A_1 has a OSPF priority 250 (to be DR)

The switch 5K_C_1 has a OSPF priority 245 (to be BDR)

The BES2 OBS is a WAN L2 link

The switch 5K_B_1 establish a FULL/DR adjanceny with 5K_A_1 (good)

 

My problème is:

The switch 5K_B_1 stay in EXSTART/BDR with the switch 5K_C_1 (the 5K_B_1 can ping the 5K_C_1).

The switch 5K_B_2 stay in EXCHANGE/DR with the switch 5K_A_1 (the 5K_B_2 can't ping the 5K_A_1).

The switch 5K_B_2 stay in EXSTART/BDR with the switch 5K_C_1 (the 5K_B_2 can't ping the 5K_C_1).

 

I think that the problem comes from that the OSPF packets/frames can't be forwarded through 5K L3 switch.

I can't succeed on adding static routes to allow the ping between the 5K that doesn't have full adjacency.

 

Any idees or suggestion are welcomes.

 

Regards Emmanuel.

 

14 Replies 14

AllezLom63194
Level 1
Level 1

I have Added static routes to allow the ping between the 5K that doesn't have full adjacency but the problem is still present.

I know that this is a Cisco Community forum and my problem is near Dell L3 switch. but I think taht the pbroblem would be the same with 2x6500 L3 switch in VSS configuration.

Maybe this explain the no responses from the Community.

Is there a moderator that can tell me if I have to remove this post ?

This is not a problem for me I can understand.

Regard, Emmanuel.

Emmanuel

 

I believe that the lack of response is not about this being a Cisco community and has more to do with a description that is difficult to understand and a lack of configuration details that might help us to understand the issue.

 

As a starting point can you verify that each of these devices is connected in the same vlan? And can you confirm that there is end to end communication. For example for devices at one end can you the the mac address of the device at the other end in the switch's mac address table? From a device at one end can you ping to the device at the other end? And does that device have an entry in its arp table for the other device (and if there is an arp entry does it have the correct mac address for the other device)?

 

Another thing to do would be to run debug ospf adjacency and post the output. (I have tried to use vendor neutral terms, but if this is a bit Cisco specific you may need to translate into your vendor terminology)

 

HTH

 

Rick

HTH

Rick

Hello Richard,

Thank you for your reply.

More details to facilitate your understand of my problem.

I was not able to ping 5K_A_1 from 5K_B_2 until I addes static route on the 5K_B_2. From now I can ping but it doesn't correct the problem.

I'm going to find if the mac address table are corrects en complete on each device and I'll give you inputs.

I have already run the "ip ospf debug" I can't specify "adgacency" in Dell. Her is the output from 5K_B_2.

11.11.11.11 is 5K_A_1.

 

Sep 23 11:12:49.968 UTC %STKUNIT0-M:CP %OSPF-5-ADJCHG: Neighbour state changed.OSPF Process 1, Nbr 11.11.11.11 on interface Vl 910 change state from 2WAY to EXSTART
Sep 23 11:12:49.854 UTC %STKUNIT0-M:CP %OSPF-5-ADJCHG: Neighbour state changed.OSPF Process 1, Sequence number mismatch trigger Nbr 11.11.11.11 on interface Vl 910 change state from EXCHANGE to 2WAY
Sep 23 11:12:09.812 UTC %STKUNIT0-M:CP %OSPF-5-ADJCHG: Neighbour state changed.OSPF Process 1, Nbr 11.11.11.11 on interface Vl 910 change state from EXSTART to EXCHANGE
Sep 23 11:12:04.910 UTC %STKUNIT0-M:CP %OSPF-5-ADJCHG: Neighbour state changed.OSPF Process 1, Nbr 11.11.11.11 on interface Vl 910 change state from 2WAY to EXSTART
Sep 23 11:12:04.806 UTC %STKUNIT0-M:CP %OSPF-5-ADJCHG: Neighbour state changed.OSPF Process 1, Sequence number mismatch trigger Nbr 11.11.11.11 on interface Vl 910 change state from EXCHANGE to 2WAY
Sep 23 11:11:38.421 UTC %STKUNIT0-M:CP %SYS-5-CONFIG_I: Configured from vty0 ( 10.45.226.147 )by cerise
Sep 23 11:11:24.668 UTC %STKUNIT0-M:CP %OSPF-5-ADJCHG: Neighbour state changed.OSPF Process 1, Nbr 11.11.11.11 on interface Vl 910 change state from EXSTART to EXCHANGE
Sep 23 11:11:19.768 UTC %STKUNIT0-M:CP %OSPF-5-ADJCHG: Neighbour state changed.OSPF Process 1, Nbr 11.11.11.11 on interface Vl 910 change state from 2WAY to EXSTART

 

I think that the problem comes from the way that I have to add a static route on 5K_B_2 next hop 5K_B_1 to reach the 5K_A_1. That mean that there is L3 connection between the 5K_A_1 et 5K_B_2 and that's could explain that they can't establish full dagacency.

 

I looking for on this way.

Regards Emmanuel.

Hello Emmanuel,

in your initial post you compare DELL VLT to Cisco VSS on Catalyst 6500.

 

I tried to look at some DELL document to understand what DELL VLT does

according to the following document

https://www.dell.com/support/article/it/it/itdhs1/how16532/how-to-set-up-virtual-link-trunking-vlt-on-dell-networking-os10?lang=en

 

It looks like that DELL VLT is more similar to Nexus vPC then to C6500 VSS.

 

The Cisco C6500 VSS feature builds a single device at the control plane from two chassis.

What happens with VSS that each pair has a single master supervisor and a single IP address per Vlan.

 

As a result of this, you can see a single OSPF adjacency forming between each VSS pair.

 

In the case of DELL VLT as your post describes the feature does not create a single logical node for all protocols.

VLT = Virtual Link Trunk appears as a way to support multi chassis LAG with LACP presenting the two member switches as a single entity at LACP level on each port channel.

This can be compared to Nexus vPC.

To be noted also Cisco Nexus do not become a single node at logical level when configuring vPC.

 

Your expectations is that all devices members of each VLT pair are able to build an OSPF adjacency.

This might be a correct expectation or not, depending on VLT implementation.

 

Stucking at Exstart / Exchange in OSPF might mean an MTU issue, but you have a working OSPF adjacency between devices 5K_B_1 and 5K_A_1.

To be noted according to your network diagram these devices are directly connected to each other via the WAN link named BES2 OBS.

Using port-channels but with a single physical link 10GE  located on those two chassis.

So there shouldn't be MTU issues on the WAN link or also this OSPF adjacency would not form.

 

Two aspects have to be considered:

a)  OSPF standard behaviour on a LAN segment

if all devices are in Vlan 910 and each member switch act as an indipendent node at OSPF level we have 6 routers on a LAN segment.

A DR /BDR election should happen and all devices should be in FULL adjacency only to DR and to BDR if the OSPF network type is the standard broadcast.

 

In cisco terms two non DR devices will not form a complete adjacency but they stop at two way state (that means I see your hellos I see my OSPF RID listed in your OSPF hello , bidirectional communication).

A two way state between non DR nodes is normal and not an issue.

This is part of OSPF standard

see

https://www.ietf.org/rfc/rfc2328.txt

 

b)  VLT implementation details

 

The two member switches are still indipendent and each of them has its own IP address in Vlan 910.

However, the only physical link is between two specific members 5K_A_1 and 5K-B_1, that are able to reach a full adjacency.

 

I will try to look again at VLT interaction with OSPF, but it might be possible that VLT feature may require direct connection between member switches to have OSPF to work.

in other terms if a second link between 5K_A_2 and 5K_B_2 would exist and added to the VLT bundle this could allow 5K_A_2 and 5K_B_2 to reach a full state in OSPF.

This is just a guess based on your notes:

 

>>

The switch 5K_B_1 stay in EXSTART/BDR with the switch 5K_C_1 (the 5K_B_1 can ping the 5K_C_1).

The switch 5K_B_2 stay in EXCHANGE/DR with the switch 5K_A_1 (the 5K_B_2 can't ping the 5K_A_1).

The switch 5K_B_2 stay in EXSTART/BDR with the switch 5K_C_1 (the 5K_B_2 can't ping the 5K_C_1).

 

Edit:

I have found the following article that describes OSPF used on multiple VLT pairs

https://weekly-geekly.github.io/articles/252763/index.html.

 

First of all, it speakes about VLT feature extensions for routing.

Also the network diagram for the case of four VLT pairs show a full mesh of physical links between all member switches.

 

This looks like a confirmation of the need to have direct links between all member switches of each VLT pair.

 

Note: this web site has some advertising over it that can trigger some firewall feature on your PC.

 

Hope to help

Giuseppe

 

 

Hello Giuseppe,

 

You are absolutly right. VLT is more similar to VPC (Nexus) because each VLT member has an IP it own address and is used to establish one port-channel to 2 physical switch.

 

The MTU mistake is effectively excluded on the BES2 WAN link but may be present on the VLT (po128) between the 2 members of the VLT domain. I'm going to look that way to.

 

Exactly I normaly should have 6 routers on à broadcast network with one DR and one BDR.

The DR is 5K_A_1 and BDR is 5K_C_1 that's why 5K_B_1 establish full adjacency with 5K_A_1 and should establish full adjacency with 5K_C_1. But it doesn't.

 

Like you I think that the problem comes from the way that 5K_B_2 is not directly connected to 5K_A_1.

On 5K_B_2, I have to add static route with next hop 5K_B_1 to ping 5K_A_1 router. that mean that there  is L3 connection between thes 2 router and that could explain the adjacency problem that believes to be on broadcast L2 network.

 

I think that I should put a L2 switch between RAD box and the two 5K_B and the other-end RAD box and the two 5K_A. This will allow a best failover situation.

 

Regards, Emmanuel.

Hello

Have you tried applying ip ospf mtu-ignore on the switches ospf enabled L3 interfaces for adjacency 


Please rate and mark as an accepted solution if you have found any of the information provided useful.
This then could assist others on these forums to find a valuable answer and broadens the community’s global network.

Kind Regards
Paul

Hello Paul,

I Don't have this command on Dell 5K switchs.

Regards, Emmanuel.

Hello


@AllezLom63194 wrote:

Hello Paul,

I Don't have this command on Dell 5K switchs.

 


Apologies I looked at the 5K and the dell didn't even register!

You have 3 VLT domains however it seems VLT1 is the Primary and it looks the only physical path between VLT 2-3 is  via VLT1?

Assuming each VLT domain has an aggregation link ( a physical interface from either switch) into VLT 1 for resiliency why then do you need full ospf adjacency between VLT 2-3 also?

 

The opsf network type then doesnt have to be broadcast it can be either point-multipoint or p2p surely thus having no DR/BDR election?


Please rate and mark as an accepted solution if you have found any of the information provided useful.
This then could assist others on these forums to find a valuable answer and broadens the community’s global network.

Kind Regards
Paul

Hello Paul,

Yes the only way To VLT2 and VLT3 is via VLT1.

The Full adjacency is for join the DR (member of VLT1 and BDR member of VLT2)

If I Don't use broadcast network, I have to setup one IP network between each VLT domain as Point_To_Point network and put these network to area 0.

 

The problem is that these L3 switch will have to communicate with other router in area 0.

 

Regards, Emmanuel.

 

Hello

Apologies then maybe I still don't understand your topology, VLT1 will be the hub for communication between VLT2-3, your only physical connections i can see is VLT3- switch--VLT1-VLT2.

 

So im not quite sure why you would wont to virtually connect  VLT2-3 to each other via ospf?


Please rate and mark as an accepted solution if you have found any of the information provided useful.
This then could assist others on these forums to find a valuable answer and broadens the community’s global network.

Kind Regards
Paul

Hello Paul,

No problem, my topology is not easy to understand and is new for me too.

 

You are right. To go from VLT2 to VLT3 I have to cross VLT1.

The RAD box is not a switch but an optical/twisted pair câble tranciever.

BES2 is L2 WAN opérator line.

 

I would like to connect VLT2 to VLT3 in ospf because VLT2 need to learn static routes redistributed by VLT3 for exemple.

 

The next topology evolution will be to add 3 Cisco ASR router on the OSPF area 0 as ABR to exchange routes with an other operator. So VLT3 will have to learn routes froms VLT2 for thes ASR.

 

Do I understand what you suggest is to define ?

 1- point to point network between VLT1 and VLT2

 2- point to point network between VLT1 and VLT3

 3- Put these 2 networks on the area 0 

 

Regards Emmanuel.

Hello


@AllezLom63194 wrote:

Hello Paul,

No problem, my topology is not easy to understand and is new for me too.

 

You are right. To go from VLT2 to VLT3 I have to cross VLT1.

The RAD box is not a switch but an optical/twisted pair câble tranciever.

BES2 is L2 WAN opérator line.

 

I would like to connect VLT2 to VLT3 in ospf because VLT2 need to learn static routes redistributed by VLT3 for exemple.


You can do this anyway in a hub spoke topology just like you have now in fact as it stands vlt2-3 could be in their own nssa area which would still rna me you to redistribute static from each belt domain 

 


Please rate and mark as an accepted solution if you have found any of the information provided useful.
This then could assist others on these forums to find a valuable answer and broadens the community’s global network.

Kind Regards
Paul

Hello Paul,

 

I can't configure hub and spoke network beacause:

1- I don't have commande ip ospf network point-to-multipoint (Dell 5K L3 switchs)

2- 5K_A_1 is not connected to 5K_B_2 nor 5K_C_1

 

I think that I have to insert two L2 switchs. One between 5K_A_x, 5K_C_x and the RAD and one between 5K_B_x and the other RAD. Then all the 5K will be on the same broadcast network and will be able to exchange each others.

 

Emmanuel.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: