Supported Platforms
Related Documentation
Understanding LACP on Chassis Clusters
Link aggregation groups (LAGs) can be established across nodes in a chassis cluster.
A redundant Ethernet interface has active and standby links located on two nodes in a chassis cluster. All active links are located on one node, and all standby links are on the other node. You can configure up to eight active links and eight standby links per node. Link aggregation allows a redundant Ethernet interface to add multiple child interfaces from both nodes and thereby create a redundant Ethernet interface LAG.
Having multiple active redundant Ethernet interface links reduces the possibility of failover. For example, when an active link is out of service, all traffic on this link is distributed to other active redundant Ethernet interface links, instead of triggering a redundant Ethernet active/standby failover.
Standalone LAG interfaces are supported on clustered devices but cannot be added to redundant Ethernet interfaces. Instead, aggregated Ethernet interfaces and redundant Ethernet interfaces coexist. (Because the functionality of a redundant Ethernet interface relies on the Junos OS aggregated Ethernet framework, you can think of it as a special aggregated Ethernet interface.) Likewise, any child interface of an existing LAG cannot be added to a redundant Ethernet interface and vice versa. The maximum number of total combined standalone aggregate interfaces (ae) and redundant Ethernet interfaces (reth) per cluster is 128.
You configure LACP on a redundant Ethernet interface by setting the LACP mode for the parent link with the lacp statement. The LACP mode can be off (the default), active, or passive.
This topic contains the following sections:
Minimum Links
Redundant Ethernet interface configuration includes a minimum-links setting that allows you to set a minimum number of physical child links in a redundant Ethernet interface LAG that must be working on the primary node for the interface to be up. The default minimum-links value is 1. When the number of physical links on the primary node in a redundant Ethernet interface falls below the minimum-links value, the interface will be down even if some links are still working.
Sub-LAGs
LACP maintains a point-to-point LAG. Any port connected to the third point is denied. However, a redundant Ethernet interface does connect to two different systems or two remote aggregated Ethernet interfaces by design. To support LACP on both redundant Ethernet interface active and standby links, a redundant Ethernet interface can be modeled to consist of two sub-LAGs, where all active links form an active sub-LAG and all standby links form a standby sub-LAG. In this model, LACP selection logic is applied and limited to one sub-LAG at a time. In this way, two redundant Ethernet interface sub-LAGs are maintained simultaneously while all the LACP advantages are preserved for each sub-LAG.
It is necessary for the switches used to connect the nodes in the cluster to have a LAG link configured and 802.3ad enabled for each LAG on both nodes so that the aggregate links will be recognized as such and correctly pass traffic.
 | Note: The redundant Ethernet interface LAG child links from each node in the chassis cluster must be connected to a different LAG at the peer devices. If a single peer switch is used to terminate the redundant Ethernet interface LAG, two separate LAGs must be used in the switch. |
Hitless Failover
With LACP, it is essential for the redundant Ethernet interface to support hitless failover between the active and standby links in normal operation. The term hitless means that the redundant Ethernet interface state remains up during failover.
The lacpd process manages both the active and standby links of the redundant Ethernet interfaces. A redundant Ethernet interface pseudolink is in the up condition when the number of active up links is not less than the number of minimum links configured. Therefore, to support hitless failover, the LACP state on the redundant Ethernet interface standby links must be collecting and distributing before failover occurs.
PDUs
By default, aggregated and redundant Ethernet links do not exchange link aggregation control protocol data units (PDUs), which contain information about the state of the link. You can configure Ethernet links to actively transmit link aggregation control PDUs, or you can configure the links to passively transmit them, sending out link aggregation control PDUs only when they receive them from the remote end of the same link. The local end of a child link is known as the actor and the remote end of the link is known as the partner. That is, the actor sends link aggregation control PDUs to its protocol partner that convey what the actor knows about its own state and that of the partner’s state.
You configure the interval at which the interfaces on the remote side of the link transmit link aggregation control PDUs by configuring the periodic statement on the interfaces on the local side. It is the configuration on the local side that specifies the behavior of the remote side. That is, the remote side transmits link aggregation control PDUs at the specified interval. The interval can be fast (every second) or slow (every 30 seconds).
By default, the actor and partner transmit link aggregation control PDUs every second. You can configure different periodic rates on active and passive interfaces. When you configure the active and passive interfaces at different rates, the transmitter honors the receiver’s rate.
