Principles

This topic describes the principles of multi-chassis link aggregation group (MC-LAG), which is an extension of the original IEEE 802.3 LAG. An MC-LAG is formed by aggregating Ethernet links between two devices with the link aggregation group (LAG) technology. The MC-LAG can provide dual-homing protection for Ethernet services.

Implementation Principles of MC-LAGs

As shown in Figure 1, an MC-LAG is made up of the following parts:

LAGs (LAG 1 and LAG 2) in dual-homing nodes. LAG 1 is the primary link to carry Ethernet services and is active currently. LAG 2 is the secondary link and is inactive currently.
LAG (LAG 3) in the router. LAG 3 interconnects with LAG 1 and LAG 2 and uses the same working link as LAG 1 and LAG 2.
Multi-chassis service paths (MCSPs). Dual-homing nodes use MCSPs to exchange LAG information and negotiate the LAG status (active or standby).

Figure 1 Dual-Homing Protection Provided by MC-LAGs

In normal situations, Ethernet services are carried by LAG 1. When a fault occurs on LAG 1, the MC-LAG automatically switches Ethernet services to LAG 2 for forwarding. In this manner, dual-homing protection for Ethernet services are implemented.

Implementation Model of MC-LAGs

The MC-LAG only supports the non-load working mode and provides protection for Ethernet links.

Requirements for LAGs are as follows:
- Support non-load and load sharing. The working mode of LAG 1 and LAG 2 must be the same.
- Support manual and static aggregation. The aggregation mode of LAG 1, LAG 2 and LAG 3 must be the same.
- Support the revertive and non-revertive modes after faults occur on LAG 1. The reversion modes of LAG 1 and LAG 2 must be the same.
Requirements for MCSPs are as follows:
To ensure fast switchover of MC-LAGs, PE 1 and PE 2 must be directly connected by an MCSP so that they can exchange LAG information with each other in MC-LAG packets.

An MCSP failure will invalidate the MC-LAG and disable it to perform active/standby switchover. In this scenario, PE 1 and PE 2 re-negotiate the working status with the Router separately and work independently based on negotiation results (like a common LAG). To improve the reliability of MC-LAGs, providing link protection (such as LAG protection and ERPS protection) for MCSPs is recommended.

Creating an MC-LAG

Create an MC-LAG in the static aggregation mode.
Exchange LACP packets between LAG 1 and LAG 3, and between LAG 2 and LAG 3 to select a link to carry services as follows:
- Based on LAG priorities and device MAC addresses, select LAG 1 or LAG 2 as the active link and the other as the standby link. The active connection should be the one with a higher priority. If the priorities of LAG 1 and LAG 2 are the same, use the one with a smaller MAC address.
- The selected LAG selects the member links to carry services based on the priorities and status of member ports after negotiating with the peer.
Create an MC-LAG in the manual aggregation mode.
NOTE:
In the manual aggregation mode, LAGs can contain only one member link.

LAG 3 cannot communicate with LAG 1 or LAG 2 in the mode. The active link of MC-LAGs must be selected based on the configurations of LAGs.
- Select the LAG on the device with a smaller MAC address to carry services. The LAG on the other device serves as the standby link.
- When configuring LAG 1, LAG 2 and LAG 3 that interconnects with LAG 1 and LAG 2, ensure that both interconnected parties use the same link to carry services.

MC-LAG Switchover

After an MC-LAG is created, the active link (supposing LAG 1) carries Ethernet services and the standby link (supposing LAG 2) does not carry any services. PE1 and PE2 exchanges MCSP packets to notify each other of LAG 1 and LAG 2 status. When a fault occurs on the active link (LAG 1), the MC-LAG sends MCSP packets to PE 1 and PE 2 and switches services to the standby link. The detailed operations are as follows:

LAGs work in the load sharing mode.
When a fault occurs on links in LAG 1, the MC-LAG compares the current working bandwidth of LAG 1 with that of LAG 2 to determine whether a switchover is needed.
- If the current working bandwidth of LAG 1 is larger than that of LAG 2, no switchover is performed and LAG 1 is still used to forward Ethernet services.
- If the current working bandwidth of LAG 1 is smaller than that of LAG 2, the MC-LAG switches Ethernet services to LAG 2 for forwarding based on the service priority.
LAGs work in the non-load sharing mode.
When a fault occurs on links in LAG 1, active/standby switchover of links is performed on the device and Ethernet services are switched to a standby link in LAG 1 for transmitting. Ethernet services are switched to LAG 2 based on service priority only when all links (active and standby) in LAG 1 fail.
NOTE:
In the manual aggregation mode, Ethernet services are switched to LAG 2 directly when a fault occurs on LAG 1. Because in this mode, LAG 1 contains only one link.