Etherchannels and the Link Aggregation Control Protocol (LACP)

The need for link aggregation

In the previous chapter, you saw how STP gives us link redundancy between switches while protecting the network from Layer 2 loops and broadcast storms. This makes the network more resilient: if one link fails, traffic can still flow over another link.

Although STP is useful, it has an important drawback if we rely on it by itself. Look at the diagram below, where two switches are connected with redundant links:

(Image 070)

Image Title - 2 Switches connected together with redundant links and STP blocking one link

STP prevents loops by blocking one of the ports on the non-root bridge (SW2). That keeps the network stable, but it also means only one physical link is actively forwarding traffic between the switches. The blocked link sits idle unless the primary link (the root port) fails.

This leads to a practical question: if we can only use one link at a time, why install multiple links? Redundancy helps with business continuity, but it would be even better if we could use all links at the same time.

Using only one link also limits bandwidth. For example:

• If both links are 1Gbps and only one is forwarding, you can send up to 1Gbps between the switches.
• If you could use both links at once, you could potentially carry 2Gbps total (1Gbps + 1Gbps).

With STP alone, you can’t actively use multiple parallel links between the same two switches without creating a loop. The solution is a technology that bundles multiple physical links into one logical link. The industry calls this Link Aggregation or LAG. Cisco calls it EtherChannel.

Etherchannels

EtherChannels solve the “one link is blocked” problem by bundling multiple physical links into a single logical interface. From STP’s perspective, that bundle looks like one link, so STP can keep the network loop-free while still allowing all member links to forward traffic.

From this point on in the book, you will see the term “interface” used interchangeably with “port”. Physical interfaces are synonymous with the word “port” in networking, (except when discussing TCP/UDP “port” numbers).

Traffic is load-balanced across the links in the bundle. That means the total available bandwidth can equal the sum of the member links. For example, if you bundle four 1Gbps interfaces into an EtherChannel, the logical EtherChannel can carry up to 4Gbps total (1Gbps + 1Gbps + 1Gbps + 1Gbps).

(Image 071)

Image Title - 2 Switches connected together with a 4 link Etherchannel. Physical and logical view

Requirements for bundling ports

To bundle ports into an EtherChannel, certain settings must match across all member ports. If they don’t match, the ports won’t bundle correctly and may even be placed into an error state.

Below is a list of the parameters that must match amongst all the ports that are members of an Etherchannel:

• Duplex (Half vs. Full)
• Port speed (100Mbps vs. 1Gbps vs. 10 Gbps etc.)
• Port mode (Access vs. Trunk)
• Native VLAN (must be the same)
• Allowed VLANs (list of VLANs allowed on the Trunk link must match)

The goal is to make multiple physical ports behave like one logical port, so their configurations must be identical.

Port-channel interfaces

The switch on the other end of the link must also be configured the same way. Once you bundle the physical ports, the switch automatically creates a logical EtherChannel interface called a Port-Channel interface.

From the switch’s perspective, the Port-Channel is a single interface. You can also create the Port-Channel interface manually first and then add physical interfaces to it later - either approach works.

After the Port-Channel exists, you should apply configuration changes to the Port-Channel interface, not to the individual member interfaces. Internally, the switch applies matching configuration to each member port. If a physical interface has settings that don’t match the Port-Channel, it can be placed into err-disabled mode.

When an interface is err-disabled, it is effectively in a “shutdown” state: it won’t send or receive frames.

You can think of an EtherChannel as a “group” of ports that you manage at the group level (the Port-Channel level). EtherChannels are also loop-free because the switch treats the bundle as one interface.

Benefits beyond bandwidth

Besides increased bandwidth, EtherChannels also provide fast self-healing. If one physical link in the bundle fails:

• The Port-Channel bandwidth is reduced (because fewer links remain).
• STP typically does not need to reconverge, because STP only sees the Port-Channel interface.

As long as the Port-Channel interface stays up, STP has no visibility into individual member links going down. For the Port-Channel interface to remain up, at least one member port must be up.

(Image 072)

Image Title - 2 Switches connected together with a 4 link Etherchannel. 2 links fail and STP does not need to converge

A key risk: mismatched configurations

A major risk with EtherChannels is inconsistent configuration between the two switches. For example, if one side uses different interface settings than the other, you can run into connectivity and performance problems.

STP can detect some situations (such as one switch bundling links while the other switch does not), but it can’t reliably detect all mismatches - for example, speed and duplex mismatches across the EtherChannel. That’s where a management protocol helps.

Link Aggregation Control Protocol (LACP)

The Link Aggregation Control Protocol (LACP) is a management protocol for EtherChannels. It dynamically manages the formation of EtherChannels and monitors their status. LACP isn’t strictly required, but it’s considered best practice.

LACP is an industry-standard link aggregation protocol defined by IEEE 802.3ad.

Sidenote

Static Etherchannels and Cisco's PAgP

Etherchannels that are configured without a management protocol are called “Static Etherchannels” and are not recommended because they are dangerous and could cause Layer 2 Loops and aren’t capable of detecting misconfigurations on the other Switch. But depending on whether the device on the other end of the Etherchannel can support LACP, this may be your only option.

Cisco developed their own proprietary Etherchannel management protocol called the “Port Aggregation Protocol” (PAgP) however, LACP is more commonly used throughout most modern deployments. PAgP is also out of the scope of the CCNA 200-301 v1.1 exam.

How LACP forms an EtherChannel

When you configure an EtherChannel, you must choose whether it will use LACP. If you configure an EtherChannel without LACP, you can’t later “convert” it to LACP - you would need to delete the Port-Channel interface and recreate it.

Once you configure LACP, the switch sends LACP packets out of the member interfaces. These packets include EtherChannel-related parameters and other LACP information used to negotiate the bundle.

• If both sides agree on the parameters, the Port-Channel interface can come up.
• If the parameters don’t match, the Port-Channel interface stays down.

This negotiation helps prevent problems caused by misconfiguration.

LACP modes: active and passive

LACP has two negotiation modes: Active and Passive.

• Active: the switch actively sends LACP packets to find and negotiate with an LACP neighbor.
• Passive: the switch waits to receive LACP packets before it responds and negotiates.

Because passive mode waits for the other side to initiate, two switches configured as Passive/Passive won’t form an EtherChannel (neither side sends LACP packets).

(Image 073)

Image Title - 2 Switches with an LACP Etherchannel. SW1 is Active and SW2 is Passive.

Switch1	Switch2	Etherchannel Formation
Active	Passive	Yes
Active	Active	Yes
Passive	Passive	No