When an inter-chassis trunk link exists between two devices, traffic sent by a network is forwarded only through local trunk member interfaces. If all local member interfaces are unavailable and the remote member interfaces are available, traffic is forwarded to the remote trunk member interfaces along inter-chassis links.

The forwarding principles are that each chassis delivers a trunk member list, in which member interfaces on the local and remote chassis are saved in different lists. In forwarding, the chassis finds local member interfaces. If there is only one local member interface and it fails, the chassis automatically selects a member interface on the other chassis. If local member interfaces are available, the chassis selects a member interface to forward packets.

Although trunk localization is used, traffic may be transmitted through different chassis. The issue occurs in the following situations:

• No trunk interface is used, or an intra-trunk interface is used.
• An inter-chassis trunk interface is used. A trunk member interface fails, or the board on which a trunk member interface resides fails.

Sufficient bandwidth resources must be reserved for traffic to be transmitted between chassis.

The bandwidth of the inter-chassis data channels can be planned both for bypass traffic if no faults occur and for maximum bypass traffic if a single point of failure occurs (for example, a board fails, causing the maximum volume of bypass traffic).

Multicast member interfaces reside on various chassis. A local chassis replicates a single copy of traffic to a remote chassis for two-level replication.

For details, see the (multicast part) forwarding plane principles of a virtual cluster.