This chapter contains a complete example of Multi-Chassis Link Aggregation (MLAG) configuration.

MLAG expands the concept of link aggregation so that it provides node-level redundancy by allowing two or more nodes to share a common LAG endpoint. It emulates multiple nodes to represent as a single logical node to the remote node running Link aggregation. As a result, even if one of the nodes is down there exists a path to reach the destination through the other nodes.

The MLAG primary and secondary physical devices synchronize with each other to create a single logical domain called Multi-Chassis Ether Channel (MCEC) domain. The exchange of information between MLAG primary and secondary devices to maintain this MCEC domain can be done with the use of IDL and IDP.

This link is used by MLAG primary and secondary devices to synchronize the states between each other via MCEC packets to create the illusion of a single control plane. The MLAG peers use the IDL to exchange control information and coordinate with each other to ensure that data traffic is forwarded properly. The IDL interface is capable of carrying data traffic:

This is an alternative way to synchronize the states between MLAG primary and secondary devices. In particular, it provides a Layer 3 communication path to carry MCEC packets. No data traffic is carried over the IDP. Hence, data traffic loss is expected, in case of MLAG interface failure.

Following are the possible types of interconnections for the IDP:

 

The IDL uses either a single physical or an aggregated ethernet interface. However, if IDL fails, the states between MLAG devices can no longer be synchronized. Configuring the IDL along with backup IDP provides more robustness during network disruptions.

The IDP without IDL can be used in cases where data traffic transmission between the primary and secondary MLAG devices are not needed.

When the IDL fails, the IDP (if configured) can be used as a backup mechanism to synchronize the states between MLAG peer devices. Also, it helps the MLAG switch to determine whether the IDL itself has failed or whether the peer MLAG device is operationally down.  When both the IDL and IDP are configured, following are the behavior of the MLAG devices when the link/device fails:

In this case, no impact on MLAG, therefore, no network downtime.

In this case, MCEC domain adjacency comes up via IDP. When the IDL link fails, MLAG Active-Active will be operational as MLAG Active-Standby (Safe Mode) to minimize network disruptions caused by Dual-Active (Split Brain) mode. It is expected that the traffic loss for hosts directly connected to the Standby device as data traffic cannot be carried over in the IDP.

When MLAG peer device reloads, MCEC domain adjacency goes down. The traffic is forwarded through the operational links in the Active MLAG device.

If both IDL and IDP link fails, Dual-Active (Split brain scenario) can be expected. There will be network instability/disruption as both MLAG nodes are operating as Active-Active with MCEC Domain Adjacency down.

In the below example TOR1 and TOR2 forms a MLAG domain.

As shown in Figure 14-31, TOR1 and TOR2 forms a domain and thus provides the redundancy. Even if TOR1 or TOR2 is down, there exist a redundant path to forward the traffic.

 

 

 

 

MLAG (also called DRNI, Distributed Resilient Network Interconnect) expands the concept of link aggregation so that it provides node-level redundancy by allowing two or more nodes to share a common LAG endpoint. MLAG emulates multiple nodes to represent as a single logical node to the remote node running link aggregation. As a result, even if one of the nodes is down there exists a path to reach the destination through the other nodes.

As shown in Figure 14-32, TOR 1 and TOR 2 form an MLAG domain. TOR 1 and TOR 2 are a single logical switch to LEAF and LwSW. Even if either TOR 1 or TOR 2 is down, there exists a path to reach other destinations.

Static MLAG provides node-level redundancy by allowing two or more nodes in the network to share a common static- LAG endpoint. It emulates multiple nodes to represent as a single logical node to the remote node having static Link aggregation. As a result, even if one of the nodes is down there exists a path to reach the destination via other nodes.