Link Aggregation and Multi-Chassis LAG
LAG and multi-chassis LAG
Link aggregation bundles parallel links into one logical link, increasing bandwidth and redundancy. Static LAGs require manual consistency; LACP (802.3ad) negotiates with active or passive modes. At least one side must be active. Member links must match speed, duplex, and VLAN/trunk configuration, and the hash distributes flows, not packets, so a single flow is limited to one member. Multi-chassis LAG on Aruba CX is VSX LAG (active-active across a VSX pair); ArubaOS-Switch uses distributed trunking. MCLAG requires a working inter-switch link and keepalive path, and it avoids STP blocking one uplink.
Traps: configuring LACP passive on both sides; mismatched member settings; expecting 2×10G to give a single 20G flow; forgetting that MCLAG depends on VSX health; and allowing STP to block MCLAG ports due to inconsistent configuration. Also check LACP system priority/MAC if the wrong side controls bundling.
Worked example/rule of thumb: A 2-member 10G LAG gives 20G aggregate, but a single TCP flow may reach only ~10G. Rule: use LACP active on both ends, keep member configs identical, and verify show lacp and VSX/MCLAG status before troubleshooting higher layers.