Cisco Nexus: vPC Cluster on Switches
Summary
The vPC 'cluster' is an approach of building multiple vPC domains (each with up to two peers), forming separate Multichassis EtherChannel (MCEC). This allows the access or distribution layer architecture to be scaled beyond two boxes, maintaining high availability, aggregating bandwidth and eliminating points of failure.
What is a vPC Cluster?
Limitation of two peers per domain
Each vPC domain supports a maximum of two switch instances (peers). You cannot have three or more devices in one vPC domain.
Multi-domain architecture
To have more than two switches, multiple vPC domains are created. Each domain combines a pair of devices into a separate MCEC. End devices (servers, top-of-rack) connect via EtherChannel to a single pair, and different domains can serve different network segments.
vPC Cluster architecture
Device pairing
Two Nexus instances in one vPC domain.
Unique domain number (e.g. 10, 20, 30)
Peer-link and Peer-keepalive
Peer-link - physical Port-Channel, trunking VLANs and carrying MCEC traffic.
Peer-keepalive - a separate channel (ICMP/BFD) that monitors the control state.
Top-of-Rack and traffic aggregation
Servers use LACP, connecting to both switches of a given pair.
Subsequent domain pairs can support other servers or VLANs
Spanning Tree and redundancy
STP treats the entire vPC domain as a single link - no loop blocking
In case of port or switch failure, traffic is taken over by the other peer.
Step by step: Configuring Multiple vPC Domains
Enable vPC on all devices
feature vpc
feature interface-vlan
Configuration of vPC Domain 10 (Peer A-B)
vpc domain 10
peer-keepalive destination 192.0.2.2 source 192.0.2.1
interface ethernet1/1-2
channel-group 100 mode active
interface port-channel100
switchport mode trunk
vpc peer-link
vPC Domain configuration 20 (Peer C-D)
vpc domain 20
peer-keepalive destination 192.0.2.4 source 192.0.2.3
interface ethernet1/3-4
channel-group 200 mode active
interface port-channel200
switchport mode trunk
vpc peer-link
3.4 Aggregation of server traffic in Domain 10
interface ethernet1/10-11
channel-group 110 mode active
interface port-channel110
switchport mode access
switchport access vlan 100
vpc 110
Aggregation of server traffic in Domain 20
interface ethernet1/12-13
channel-group 210 mode active
interface port-channel210
switchport mode access
switchport access vlan 200
vpc 210
Verification of the status of all domains
show vpc brief
show vpc consistency-parameters
Best practice
Each vPC domain must have a unique number.
Port-channel parameters (VLAN, MTU, trunk/access) must be identical on both sides.
Peer-link should consist of at least two physical links for redundancy.
Regular monitoring with `show vpc` commands and log analysis.
Applications
Data centres - redundancy and aggregation of server connections.
Distribution layer - scalability over two boxes.
Critical applications - minimising points of failure and high availability.
Sources