Questions to Determine¶
- What LACP mode is configured on the server bond (active vs passive)?
- What LACP mode is configured on the switch port-channel (active vs passive)?
- Are both sides actually negotiating LACP, or has one side fallen back to static aggregation?
- What does
cat /proc/net/bonding/bond0show for each slave's LACP partner info? - What hashing algorithm is the switch using for load distribution across the port-channel members?
- Does the switch port-channel config match the server bond config (e.g., is the switch using src-mac only)?
- Are the per-link traffic counters confirming the imbalance?
- What is the LACP rate (slow/fast) on each side?
- Could the traffic pattern (e.g., single large flow between two IPs) inherently hash to one link?
- Are there any LACP error counters or negotiation failures logged?