Enable OSPF between CE and PE using EVPN Anycast

[Mas627]

I am testing EVPN Anycast on IOS XE.  I have it working successfully, but when I enable OSPF between the CE, PE1 and PE2 the OSPF sessions constantly flap.  Is this configuration not supported?  I guess because the CE and PEs are in the same EVPN instance it sees the Anycast address as a duplicate from two different router IDs?

[M02@rt37]

Hello @Mas627 

It's possible that the issue you're experiencing is due to the EVPN Anycast address being seen as a duplicate from two different router IDs, as you suggested.

When using EVPN Anycast, the Anycast address is shared by multiple PEs, and it is advertised to the CE as a reachable next-hop for the MAC addresses being advertised via EVPN. However, when OSPF is also enabled between the CE and the PEs, the PEs will also advertise the Anycast address as a reachable next-hop for OSPF routes, and this can/could cause the OSPF sessions to flap as each router sees the Anycast address as coming from a different router ID.

You can try to configure the PEs to advertise the Anycast address as a type-2 or type-3 external LSAs in OSPF, instead of as a reachable next-hop for OSPF routes. This way, the CE will learn the Anycast address as an external route, and it won't be seen as a duplicate from different router IDs.

[Mas627]

Thanks for the info! How do you advertise the anycast as type-2 or 3? Are you referring to configuring the interface as P2P? 

[sidshas03]

When advertising an anycast address in a routing protocol, the type of network that the anycast address is associated with is determined by the type of network configured on the interface that is advertising the anycast address.

In OSPF, for example, the network type is determined by the

OSPF network command

used to configure the interface. To advertise an anycast address as a Type 2 or Type 3 network, you would configure the interface as a Point-to-Multipoint (P2MP) network type.

For example, in OSPF, you would configure the interface as follows:

interface gigabitethernet 0/0
 ip address 10.0.0.1 255.255.255.0
 ip ospf network point-to-multipoint

In this example, the network type is set to P2MP using the

ip ospf network point-to-multipoint command

This will cause the anycast address associated with this interface to be advertised as a Type 2 or Type 3 network, depending on the OSPF area configuration.

It's important to note that the specific commands and syntax may vary depending on the routing protocol being used and the specific network topology.

[M02@rt37]

@Mas627 

To configure the PEs to advertise the Anycast address as a type-2 or type-3 external LSA in OSPF, you can do that:
=> Determine the Anycast address that is being used for EVPN Anycast on your network. This address should be the same across all PEs that are participating in the EVPN instance.
=> On each PE, configure a loopback interface with the Anycast address. This loopback interface will be used to advertise the Anycast address in OSPF.

For example:

interface Loopback0
ip address <anycast-address>/32
ip router ospf 1 area 0

Configure the PEs to redistribute the loopback interface into OSPF as an external LSA. You can use the redistribute command under the OSPF process to do this. Here is an example configuration for redistributing the loopback interface as a type-3 external LSA:

router ospf 1
redistribute connected subnets metric-type 3

[Mas627]

But then how does CE1 establish an adjancency to PE1 and PE2? 

CE1 has interface address of 192.168.10.1 and PE have Anycast addres of 192.168.10.254.  Moving the anycast address to a loopback means CE1 won’t be able to establish OSPF neighbour . 

[M02@rt37]

@Mas627 

You're correct that moving the anycast address to a loopback interface on the PEs would prevent the CE from establishing an OSPF adjacency.

Then, you could use a different router ID on each of the PEs when advertising the anycast address in OSPF. For example, on PE1 you could use router ID 1.1.1.1 and on PE2 you could use router ID 2.2.2.2. Then, when the CE receives OSPF updates from the PEs, it will see the anycast address as coming from two different router IDs and will not consider them duplicates.

[Mas627]

M02@rt37  this is what I did with the intial configuration and seen OSPF adjancency flapping.  I followed @sidshas03 configuring P2MP and this seems to have resolved the issue.  I am doing more testing to make sure it's all okay. 

Thanks both for your assistance!

[Mas627]

M02@rt37  @sidshas03

Unfortunately the issue still persists, after setting ospf interfaces P2MP OSPF still flaps.  When I shutdown the link between CE1 and PE2, OSPF goes into FULL and stays stable. 

Maybe need to check with TAC on this?

output from CE1

*Mar 13 10:28:24.781: %OSPF-5-ADJCHG: Process 1, Nbr 2.2.2.2 on Vlan10 from EXSTART to DOWN, Neighbor Down: Too many retransmissions
*Mar 13 10:29:24.781: %OSPF-5-ADJCHG: Process 1, Nbr 2.2.2.2 on Vlan10 from DOWN to DOWN, Neighbor Down: Ignore timer expired
*Mar 13 10:31:16.707: %OSPF-5-ADJCHG: Process 1, Nbr 1.1.1.1 on Vlan10 from EXSTART to DOWN, Neighbor Down: Too many retransmissions
*Mar 13 10:31:47.861: %OSPF-5-ADJCHG: Process 1, Nbr 2.2.2.2 on Vlan10 from EXSTART to DOWN, Neighbor Down: Too many retransmissions
*Mar 13 10:32:16.707: %OSPF-5-ADJCHG: Process 1, Nbr 1.1.1.1 on Vlan10 from DOWN to DOWN, Neighbor Down: Ignore timer expired
*Mar 13 10:32:47.861: %OSPF-5-ADJCHG: Process 1, Nbr 2.2.2.2 on Vlan10 from DOWN to DOWN, Neighbor Down: Ignore timer expired
*Mar 13 10:34:38.674: %OSPF-5-ADJCHG: Process 1, Nbr 1.1.1.1 on Vlan10 from EXSTART to DOWN, Neighbor Down: Too many retransmissions
*Mar 13 10:35:08.310: %OSPF-5-ADJCHG: Process 1, Nbr 2.2.2.2 on Vlan10 from EXSTART to DOWN, Neighbor Down: Too many retransmissions
*Mar 13 10:35:38.674: %OSPF-5-ADJCHG: Process 1, Nbr 1.1.1.1 on Vlan10 from DOWN to DOWN, Neighbor Down: Ignore timer expired
*Mar 13 10:36:08.310: %OSPF-5-ADJCHG: Process 1, Nbr 2.2.2.2 on Vlan10 from DOWN to DOWN, Neighbor Down: Ignore timer expired
*Mar 13 10:37:57.114: %OSPF-5-ADJCHG: Process 1, Nbr 1.1.1.1 on Vlan10 from EXSTART to DOWN, Neighbor Down: Too many retransmissions
*Mar 13 10:38:29.596: %OSPF-5-ADJCHG: Process 1, Nbr 2.2.2.2 on Vlan10 from EXSTART to DOWN, Neighbor Down: Too many retransmissions
*Mar 13 10:38:57.114: %OSPF-5-ADJCHG: Process 1, Nbr 1.1.1.1 on Vlan10 from DOWN to DOWN, Neighbor Down: Ignore timer expired
*Mar 13 10:39:29.596: %OSPF-5-ADJCHG: Process 1, Nbr 2.2.2.2 on Vlan10 from DOWN to DOWN, Neighbor Down: Ignore timer expired
*Mar 13 10:41:17.697: %OSPF-5-ADJCHG: Process 1, Nbr 1.1.1.1 on Vlan10 from EXSTART to DOWN, Neighbor Down: Too many retransmissions
*Mar 13 10:41:51.765: %OSPF-5-ADJCHG: Process 1, Nbr 2.2.2.2 on Vlan10 from EXSTART to DOWN, Neighbor Down: Too many retransmissions
*Mar 13 10:42:17.697: %OSPF-5-ADJCHG: Process 1, Nbr 1.1.1.1 on Vlan10 from DOWN to DOWN, Neighbor Down: Ignore timer expired

[sidshas03]

Based on the provided output, it seems that there are issues with OSPF neighbor adjacency formation. The error message "Neighbor Down: Too many retransmissions" indicates that the OSPF neighbors are not able to establish a stable adjacency due to excessive retransmissions. This could be caused by various factors, such as network congestion, misconfigured OSPF parameters, or hardware issues. 

You mentioned that you have already set the OSPF interfaces to P2MP OSPF. This should help in reducing the number of OSPF adjacencies, but it seems like it is not working as expected.

It's always a good idea to open a TAC case when you're facing issues with network devices. They can help you troubleshoot the issue and provide you with a solution.

https://www.cisco.com/c/en/us/support/docs/ip/open-shortest-path-first-ospf/13699-29.html

https://www.cisco.com/c/en/us/support/docs/ip/open-shortest-path-first-ospf/12151-trouble-main.html

[rais]

What data plane is being used? EVPN CE running OSPF should become adjacent to another intra-area CE (e.g. on a p2p) on the same EVPN.

[Mas627]

The issue in my scenario I need routing between CE and PE, as the global routing table needs to learn CE subnets for internet access.  not sure what you mean by data plane, it's just OSPF and BGP being used. 

[Mas627 · Accepted solution]

Spoke to our Cisco partner, confirmed OSPF won't work but can use static routes and redistribute it to the rest of the network.  Thanks again for the help.