Eth-Trunk Static and LACP

Eth-Trunk Static and LACP

Quick Answer: Huawei Eth-Trunk aggregates multiple physical ports into one logical link. Use manual mode when both ends are forced and simple, and LACP (mode lacp-static) when you want negotiation of active members between switches. Traffic load-shares by hash across members; STP/RSTP/MSTP sees the bundle as a single link, so aggregation increases capacity without creating a Layer 2 loop between the same neighbors.

Spanning tree intentionally blocks redundant paths that form loops. Link aggregation is different: multiple cables between the same two devices are combined into one logical interface. From STP’s point of view there is only one link, so all members can forward in the active topology (subject to LACP selection). That is how you get both redundancy and more bandwidth between access and aggregation, or between aggregation and core.

Problems Eth-Trunk solves

Eth-Trunk does not replace STP across a mesh of many switches; it complements STP on parallel cables between a pair of devices (or between a switch and a dual-homed server NIC team that supports LACP).

Logical interface model on Huawei

You create a logical interface and join physical ports as members:

[SW] interface Eth-Trunk 1
[SW-Eth-Trunk1] port link-type trunk
[SW-Eth-Trunk1] port trunk allow-pass vlan 10 20 30

Then add members—two common styles exist depending on software style and habit:

[SW] interface GigabitEthernet0/0/1
[SW-GigabitEthernet0/0/1] eth-trunk 1
[SW] interface GigabitEthernet0/0/2
[SW-GigabitEthernet0/0/2] eth-trunk 1

or using trunkport-style membership from the Eth-Trunk view on some platforms/documentation patterns:

[SW] interface Eth-Trunk 1
[SW-Eth-Trunk1] trunkport GigabitEthernet 0/0/1 to 0/0/2

HCIA questions care that members belong to the Eth-Trunk, that Layer 2 attributes are configured on the Eth-Trunk, and that physical members should not carry conflicting independent VLAN configs that fight the bundle.

Working modes: manual vs LACP

Huawei documents several aggregation modes; for HCIA-Datacom focus on these two ideas:

Manual / static load-balance mode

• Members are administratively forced into the bundle.
• No LACP PDUs negotiate capability with the peer.
• Both ends must be configured correctly; a mismatch (one side bundled, the other not) can cause loops or black holes because STP may not see separate physical links the way you expect if only one side aggregates.
• Useful for simple lab setups or devices that cannot speak LACP, but weaker for production switch-to-switch links.

Conceptual configuration pattern:

[SW] interface Eth-Trunk 1
[SW-Eth-Trunk1] mode manual load-balance

(Exact mode keyword variants appear in study materials as manual / static load-balance—know the behavior: no LACP negotiation.)

LACP static mode (lacp-static)

• Uses Link Aggregation Control Protocol (IEEE 802.3ad/802.1AX family concepts).
• Peers exchange LACPDUs to agree which members are active.
• Detects member failures and misconfigurations more safely than pure manual mode.
• Recommended for switch-to-switch aggregation on Huawei campus designs at HCIA level.

[SW] interface Eth-Trunk 1
[SW-Eth-Trunk1] mode lacp-static

Exam trap: “Static” in lacp-static still uses LACP PDUs between configured peers; it is not the same as fully non-negotiating manual mode. Do not confuse “static LACP” with “no protocol.”

Member link rules and common failures

For stable aggregation:

1. Same speed and duplex on members that should be active together.
2. Same link type / VLAN allowance applied via the Eth-Trunk logical interface.
3. Both ends configured for aggregation (and the same mode family).
4. Physical members not left as independent access ports with different PVIDs while also joined to the trunk.

If member speeds differ, LACP-oriented implementations typically exclude lower-speed links so only the highest consistent speed set remains active—this avoids unpredictable hashing across unequal pipes. Manual mode may be even less forgiving if you force mismatched links. Exam items often test the idea that speed mismatch prevents unequal members from load-sharing as equals.

Maximum member counts and whether members can span different cards depend on platform; HCIA rarely needs a specific chassis limit, but you should know aggregation is subject to device hardware rules.

Load balancing concepts

Eth-Trunk does not strip one TCP flow into byte-by-byte striping across all cables in the general campus case. Instead, a hash of header fields maps each flow to one member:

• Possible inputs (platform-dependent): source/destination MAC, source/destination IP, TCP/UDP ports, or combinations.
• Many concurrent flows → good utilization across members.
• One elephant flow → can pin to a single member and still congest that link even if other members are idle.

You may see commands conceptually like setting a load-balance profile on the Eth-Trunk or globally; remember the principle more than memorizing every hash keyword. For troubleshooting “only one fiber is hot,” check flow diversity and hash fields before assuming a member is broken.

Interaction with STP and VLANs

• Configure STP cost and edge settings on the Eth-Trunk, not inconsistently on each member.
• The bundle is one STP port: either the whole logical link is root/designated/alternate for that instance, or not.
• VLAN trunking (port link-type trunk, allowed VLAN lists) belongs on the Eth-Trunk interface so all members carry the same VLAN set.
• When dual-homing an access switch to two aggregation switches, you typically use two separate uplinks or two Eth-Trunks (one toward each upstream device), not one Eth-Trunk split across two different chassis—unless those chassis are stacked/virtualized into one logical device (next section).

Verification

Primary command:

[SW] display eth-trunk 1

Interpret:

• Working mode (manual vs LACP).
• Which physical interfaces are members.
• Which members are Selected / Active vs unselected.
• Number of ports in up state and operational status of the logical link.

Also useful:

[SW] display interface Eth-Trunk 1
[SW] display lacp statistics eth-trunk 1

If members show up physically (display interface brief) but are not Selected, suspect mode mismatch, LACP peer issues, speed mismatch, or local configuration errors.

Worked design vignette

Access switch ASW dual-homes to aggregation AG1 with two GE fibers. Goals: 2 Gbit/s aggregate capacity for many users and survive one fiber cut.

1. Create Eth-Trunk 10 on ASW and AG1.
2. Set mode lacp-static on both ends.
3. Join GE0/0/1 and GE0/0/2 to Eth-Trunk 10 on both ends.
4. Make Eth-Trunk 10 a trunk allowing the user VLANs.
5. Ensure STP root still sits on core/aggregation as designed; the Eth-Trunk is simply one uplink with lower effective congestion.
6. Verify both members Selected; pull one fiber and confirm the logical link stays up and traffic continues on the survivor.

If someone instead connected the two fibers to two different aggregation switches without stacking/M-LAG-style designs, they must not place both fibers in one Eth-Trunk toward “two peers.” That is a different topology using STP (or advanced multi-chassis features beyond basic HCIA Eth-Trunk scope).

Exam checklist

• Define Eth-Trunk purpose: bandwidth + redundancy as one logical link.
• Contrast manual (no LACP) vs lacp-static (LACP negotiation).
• Know member consistency rules and speed-mismatch behavior.
• Explain hash-based load sharing and single-flow limits.
• Recall interface Eth-Trunk, membership commands, and display eth-trunk.

Aggregation between two physical switches is powerful; stacking makes multiple switches themselves behave as one logical switch so multi-chassis designs become simpler at the access layer.