cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
809
Views
0
Helpful
5
Replies

vPC fabric peering issue

shados
Level 1
Level 1

Hi, I have the topology with two orphan devices that are on different vlans but vpc peers would never seem to sync arp over cfsoip and hence the ping from one orphan device to another isn't working. Is it expected scenario? Config attached. The SW1 has no play in this, disregard it please.

shados_0-1745762475658.png

shados_1-1745762510135.png

 

2 Accepted Solutions

Accepted Solutions

As per configuration guide

https://www.cisco.com/c/en/us/td/docs/dcn/nx-os/nexus9000/102x/configuration/vxlan/cisco-nexus-9000-series-nx-os-vxlan-configuration-guide-release-102x/m_configuring_vpc_fabric_peering_93x.html

 

TCAM should be carved in specific way for fabric peering. It seems there is no such lines in your configuration.

Also it could be issue with virtualization, as NX-OSv doesn't support all features of devices. Most common example - issues with multicast.

View solution in original post

shados
Level 1
Level 1

everything seems to be fine when working with actual hardware, thanks!

View solution in original post

5 Replies 5

Pavel Tarakanov
Cisco Employee
Cisco Employee

According to provided config, virtual peer-link used instead of physical one. In such case such interaction (between hosts connected with orphan ports) more like interaction between two independent VTEPs, not as two orphan ports in classical VPC domain with physical peer link.

 

I'd recommend to check all related tables (MAC address, l2 routes, BGP L2VPN EVPN, routing tables of corresponding VRF).

Also can H1 ping to H2? Can H1 and H2 ping their default gateways?

 

Additionally, how you get this incomplete record for 192.168.20.11 on LEAF-1? Most likely you tried to ping H2 from VTEP LEAF-1. But in such case source for such ARP request will be SVI for VLAN 20 with IP address shared between all VTEPs. And ARP reply could be consumed be LEAF-2.

Hi Pavel,

Ping from H1 to H2 fails. 
I can ping the gateways:

shados_1-1745779293853.png

shados_0-1745779261618.png

I can see the MAC and IP addresses in l2rib. Meaning they are advertised and imported correctly

shados_6-1745779452955.png

shados_7-1745779484201.png

shados_5-1745779433321.png

shados_9-1745779595538.png

 

And I see the pings on the link among spines and leaves but these icmp requests are never being decapsulated and forwarded to an endpoints.

shados_8-1745779521423.png

 

 

 

 

 

 

 

As per configuration guide

https://www.cisco.com/c/en/us/td/docs/dcn/nx-os/nexus9000/102x/configuration/vxlan/cisco-nexus-9000-series-nx-os-vxlan-configuration-guide-release-102x/m_configuring_vpc_fabric_peering_93x.html

 

TCAM should be carved in specific way for fabric peering. It seems there is no such lines in your configuration.

Also it could be issue with virtualization, as NX-OSv doesn't support all features of devices. Most common example - issues with multicast.

I will try this on the hardware, I only tried it on the n9kv and the mentioned tcam carving is not available. Thanks!

shados
Level 1
Level 1

everything seems to be fine when working with actual hardware, thanks!

Review Cisco Networking for a $25 gift card