Description

This article describes the issue of preemption not being supported in multichassis link aggregation (MC-LAG) when running in Active-Standby mode (prior to 12.3 and later releases).

Symptoms

Consider an MC-LAG scenario in which one link is active and the other is in standby mode (also known as Active-Standby mode). For more information, refer to the following links:

Note: When the active link fails, it switches to the standby link. When the active link is reestablished, it is not automatically reverted.

Initial state - on the active node:

root@MX240-16-re0# run show interfaces mc-ae
Member Link : ae0
Current State Machine's State: mcae active state
Local Status : active <
Local State : up
Peer Status : standby
Peer State : up
Logical Interface : ae0.0
Topology Type : NULL
Local State : up
Peer State : Unknown
Peer Ip/MCP/State : Unknown

Initial state - on the standby node:

root@MX960-138-RE1# run show interfaces mc-ae
Member Link : ae0
Current State Machine's State: mcae standby state
Local Status : standby <
Local State : up
Peer Status : active
Peer State : up
Logical Interface : ae0.0
Topology Type : NULL
Local State : up
Peer State : Unknown
Peer Ip/MCP/State : Unknown

Active link fails (here, the failure of the active link is simulated by disabling AE on the active node):

[edit]
root@MX240-16-re0# set interfaces ae0 disable

[edit]
root@MX240-16-re0# commit

commit complete

On the initial active node:

root@MX240-16-re0# run show interfaces mc-ae
Member Link : ae0
Current State Machine's State: mcae standby state
Local Status : standby <
Local State : down
Peer Status : active
Peer State : up
Logical Interface : ae0.0
Topology Type : NULL
Local State : up
Peer State : Unknown
Peer Ip/MCP/State : Unknown


root@MX960-138-RE1# run show interfaces mc-ae
Member Link : ae0
Current State Machine's State: mcae active state
Local Status : active <
Local State : up
Peer Status : standby
Peer State : down
Logical Interface : ae0.0
Topology Type : NULL
Local State : up
Peer State : Unknown
Peer Ip/MCP/State : Unknown

Now the active link is reestablished:

[edit]
root@MX240-16-re0# delete interfaces ae0 disable
[edit]
root@MX240-16-re0# commit
commit complete

On the initial active node:

root@MX240-16-re0# run show interfaces mc-ae
Member Link : ae0
Current State Machine's State: mcae standby state
Local Status : standby < still in standby. not reverted.
Local State : up
Peer Status : active
Peer State : up
Logical Interface : ae0.0
Topology Type : NULL
Local State : up
Peer State : Unknown
Peer Ip/MCP/State : Unknown

On the standby node:

root@MX960-138-RE1# run show interfaces mc-ae
Member Link : ae0
Current State Machine's State: mcae active state
Local Status : active < still standby. not reverted.
Local State : up
Peer Status : standby
Peer State : up
Logical Interface : ae0.0
Topology Type : NULL
Local State : up
Peer State : Unknown
Peer Ip/MCP/State : Unknown

Solution

This behavior, as seen in the above example, is expected behavior as preemption is not supported with the MC-LAG feature on Junos OS releases prior to 12.3.

Configuration support for manual and automatic link switchover mechanisms on the multichassis link aggregation interface  is mentioned in the 12.3 Release Notes, New Features, Layer 2 section:

http://www.juniper.net/techpubs/en_US/junos12.3/information-products/topic-collections/release-notes/12.3/index.html?topic-69609.html#jd0e9311

In the 13.3 docs and later, configuring revertive or nonrevertive mode is discussed here:

http://www.juniper.net/techpubs/en_US/junos13.3/topics/task/configuration/interfaces-configuring-mc-lag-link-switchover.html