MCT Failover Scenarios

The following scenarios describe what happens if specific elements in the MCT configuration fail:

• Client interface on one of the MCT cluster devices goes down.
  • Traffic switches to the other cluster device with minimal traffic loss.

• MCT cluster device goes down.
  • When an MCT cluster device goes down (for example, due to a power failure), the traffic fails over to the other MCT cluster device.
  • When an ICX stack is configured as an MCT cluster and a stack member goes down, traffic converges again within 10 to 30 seconds.
    Note: Hitless failover within an ICX stack is enabled by default. Hitless failover and how it is configured for an MCT system is described separately in the following points.

• Hitless failover occurs.
  • The MCT CCEPs stay up during hitless switchover, failover, or upgrade. Link protocols such as UDLD and LACP on CCEPs do not flap. Traffic disruption is minimal (sub-second). The MCT CCP connection flaps once, and MAC is re-synced between the peer devices.
  • The CCP goes down and comes back up again once the hitless failover is completed.

• ICL interface or CCP goes down (keep-alive VLAN is configured).
  • If a keep-alive VLAN is used, the devices in the cluster can communicate even if the ICL goes down. If the peer device is reachable over the keep-alive VLAN, the MCT peers perform the active/standby negotiation per client. After negotiation, the standby shuts down its client ports, and the active client ports continue to forward the traffic.
  • The active/standby negotiation is performed per MCT client on the basis of RBridge ID and client local or remote accessibility. If the client is reachable from both MCT devices, the lower RBridge ID becomes the active device. If the client can be accessed only from one of the MCT devices, the cluster device on which it is reachable becomes the active device.
  • If the peer device cannot be reached over the keep-alive VLAN, then both cluster devices keep forwarding.

• ICL interface or CCP goes down (keep-alive VLAN is not configured).
  • When the keep-alive VLAN is not configured, both cluster devices keep forwarding. Use the client-isolation strict command to disable the client interface as soon as the ICL link goes down to completely isolate the client.

• Double failures occur (for example, the ICL goes down and the client interface goes down on one of the MCT cluster devices).
  • Multiple failures could cause traffic to drop, even if there is a physical path available.