Configuring additional MCT cluster parameters

Peer Keepalive

The peer-keepalive feature helps in distinguishing interface failure and cluster peer node failure, and takes action accordingly. An out-of-band keep-alive session is recommended between the nodes. This helps to distinguish peer-interface failures from peer node reboot. Different actions can then be taken for these peer-interface failures.

The keep-alive session has 2 flavors:

Auto keep-alive session (default)
- Does not require any explicit user configuration, and is run in the management VRF. The management IPs of the two nodes are exchanged once the cluster session is established and an out-of-band keep-alive session is started between them.
Manual keep-alive session
- A source interface with IP and destination IP can be specified to be used for keep-alive session. This can be in any VRF. These IP addresses cannot match the peer IP and source IP used for the CCP session to ensure proper out-of-band functionality.

The peer-keepalive role determines the node‘s behavior when the ICL goes down, but the MCT peer remains up.

Use auto to assign roles automatically. By default, node with lower IP is selected as primary and higher IP is selected as secondary. This behavior can be changed by manually assigning the primary/secondary role under the cluster.
The primary role is responsible for forwarding all traffic when the ICL goes down. It keeps all the cluster-client ports up and becomes the Designated Forwarder for all VLANs/BDs on the tunnels and PWs.
The secondary role isolates the secondary node, brings down all client interfaces and tunnels, and isolates all client traffic. this behavior is only effective when the keep-alive session is up while CCP session goes down. In all other cases, primary and secondary will not have an impact on functionality.

When a keep-alive session goes down but the CCP session remains up, the cluster continues to operate in regular mode. No traffic flow change is observed. This mode is not recommended because there is no peer-link protection. If the CCP session were also to go down, the behavior would be same as a peer reload; both nodes would keep the clients up and become the DF for all VLANs/BDs.

Use the peer-keepalive command to configure the peer-keepalive mode on a node, as shown in the following example.

device(config)# cluster leaf1_2
device(config-cluster-leaf1_2)# peer-keepalive
device(config-peer-keepalive)# role primary
device(config-peer-keepalive)#

Configuring peer-keepalive destination

Use the peer-keepalive destination ip command to overwrite the default keepalive session with the configured session between the specified source and the destination IP address

device(config)# cluster leaf1_2
device(config-cluster-leaf1_2)# peer-keepalive destination 10.20.161.158
device(config-cluster-leaf1_2)#

Moving the traffic from an MCT node to the remote node

Use the shutdown {all | clients} command to move all the traffic on the node to the remote MCT node by disabling the local client interfaces administratively. Use of the all parameter will shutdown the MCT peer and clients.

device(config)# cluster leaf1_2
device(config-cluster-leaf1_2)# shutdown clients
device(config-cluster-leaf1_2)#