Description

When one of the virtual chassis (VC) members in a two-member virtual chassis system is rebooted and if there is any BUM traffic passing through the other member/members in the same VC, you may see looped BUM traffic for a short while. This article provides more details on the behavior and provides a workaround to prevent the loop.

 

Symptoms

The above behavior would be seen when the following conditions are met:

  • There are two or more members in a virtual chassis system, which supports the fast failover feature.

  • BUM (Broadcast, Unknown Unicast, Multicast) traffic is passing through the remaining member/members.

  • Virtual Chassis (VC) cables are connected in a ring topology.

  • Dedicated VC ports or VCEP with fast failover is enabled.

 

Solution

Looped BUM traffic is created by the VC cables. This is due to the fast failover feature that is enabled for faster VC convergence. By default, the fast failover option is enabled on dedicated VCP ports as shown below: 

root@ex4550VC# run show virtual-chassis fast-failover

Fast failover on dedicated VCP ports: Enabled
Fast failover on XE uplink VCP ports: Disabled
Fast failover on GE uplink VCP ports: Disabled
Fast failover on ET uplink VCP ports: Disabled

When you are using Virtual Chassis Extended Ports (VCEP) with the fast failover option enabled via the following command, you may see this behavior. 

set virtual-chassis fast-failover xe

root@ex4550VC# run show virtual-chassis fast-failover

Fast failover on dedicated VCP ports: Enabled
Fast failover on XE uplink VCP ports: Enabled
Fast failover on GE uplink VCP ports: Disabled
Fast failover on ET uplink VCP ports: Disabled

When one of the VC members in this setup is rebooted and a link-down event ensues, the ports will be put in loopback mode. This is expected behavior.

However, on a two-member virtual chassis system, this would cause a short loop before both the VCP ports go down. Then looped BUM traffic would be observed on the device that is connected to the remaining member/members. This is a limitation on legacy EX Series switches and can be seen with a VC ring topology, which has two or more members in the VC.

 

On two-member virtual chassis systems that are configured with a dedicated VCP, disable the fast failover feature by using the following command since it is not required in a two-member virtual chassis setup. 

set virtual-chassis fast-failover vcp disable

If you are using VCEP and have enabled the fast-failover option, then remove the fast-failover configuration.

If you are using a VC with three or more members in a ring topology, BUM traffic loop for a short while is expected behavior when the fast-failover function is enabled. No action is required in this case because the traffic loop will be cleared automatically.