Grading Checklist¶
A good response must include:
- Identifies the root cause: LACP hashing mismatch between server and switch, causing traffic imbalance
- Explains that the switch is using a different hashing algorithm (e.g., src-mac) than the server (layer3+4), resulting in all return traffic hashing to one link
- Shows how to inspect LACP state on the server using
/proc/net/bonding/bond0 - Identifies the per-link traffic stats showing the imbalance
- Explains that both links being UP does not mean traffic is balanced -- LACP does not guarantee equal distribution
- Explains how LACP hashing works: each side independently decides which link to send traffic on
- Recommends aligning the hashing algorithm on both sides (e.g., both using layer3+4 / src-dst-ip-port)
- Mentions that single large flows cannot be split across links (flow-level, not packet-level balancing)
- Suggests reviewing the switch port-channel configuration for the hashing policy
- Considers whether the traffic pattern itself is skewed (e.g., one dominant client IP)
- Does NOT suggest simply adding more links without fixing the hashing