EtherChannel

By Alec Gauldin, Principal Network Engineer · Updated July 2026

EtherChannel (also called a port channel) aggregates as many as eight active physical links into a single logical link between two devices. You gain three things: more usable bandwidth, redundancy — if one member fails the channel stays up on the survivors — and clean interaction with Spanning Tree. Because STP/RSTP see the whole bundle as one logical link, they never put the redundant members into a blocking state, so you get every link's throughput and a loop-free topology at the same time. That combination is exactly why EtherChannel is an explicit CCNA 200-301 exam topic (roughly item 2.4, 'Configure and verify Layer 2 EtherChannel with LACP'); treat any published weighting as approximate, since Cisco adjusts it between blueprint revisions.

There are three ways to form a bundle, and the difference is negotiation. LACP (IEEE 802.3ad, later folded into 802.1AX) is the open, vendor-neutral standard and uses the modes active and passive — at least one side must be active, so active/active and active/passive both form while passive/passive does not. PAgP is Cisco-proprietary and uses desirable and auto, where desirable/desirable and desirable/auto form but auto/auto does not. Static 'on' mode runs no negotiation at all and must be set identically on both ends; because it skips the consistency handshake, a mismatch can create a loop, which is why LACP is the recommended default. The 'Configure EtherChannel with LACP' guide walks the full active-mode build.

EtherChannel comes in two flavors. A Layer 2 EtherChannel carries VLANs as an access or trunk port and is the CCNA-core case. A Layer 3 (routed) EtherChannel instead carries an IP address on the logical Port-channel interface after you apply 'no switchport' to the members and the port-channel — the pattern the 'Layer 3 (Routed) EtherChannel' guide covers, typically to bond uplinks between routing devices. In both cases you build the bundle by adding member interfaces to a channel-group; the Port-channel interface is created automatically, and that logical interface — not the individual members — is where you apply switchport/trunk settings or the IP address.

Consistency is what makes or breaks a bundle. Every member port must agree on speed and duplex, and for a Layer 2 channel they must also match switch mode (access vs trunk), native VLAN, and the allowed-VLAN list. A port whose settings disagree is suspended or left standalone and will not join the channel. The practical habit is to configure the Port-channel interface itself so the members inherit consistent settings, rather than tweaking each physical port by hand.

Load balancing decides which member a given frame rides. The switch hashes selected fields — set by 'port-channel load-balance' to source/destination MAC, IP, or (on capable platforms) TCP/UDP port — and the default input varies by platform. Distribution is per flow, not per packet, so a single conversation always uses one link and never exceeds one member's speed; you scale throughput across many flows, not within one. For even hashing, use a power-of-two number of links (2, 4, or 8). Verify the whole thing with 'show etherchannel summary', where the P flag means a port is bundled in the channel.

To master EtherChannel, work the layers in order: understand this map first, then keep the cheat sheet handy for the channel-group modes and the show commands, and finally build and grade the labs so the configuration and the verification output become muscle memory. Start with the Layer 2 LACP path (the CCNA target), confirm you can read 'show etherchannel summary' flags, then extend into routed EtherChannel and load-balance tuning.