LACP, VLAN, and Jumbo Frame Failures on a DD6400 Backup Network
A DD6400 is connected to two top-of-rack switches. Data ports eth2 and eth3 form bond0 for the backup VLAN 120, and eth4 and eth5 form bond1 for replication VLAN 220. ethMa is on the management VLAN. The DD interfaces are configured with MTU 9000, LACP active, and VLAN tagging. Backups from 200 VMware clients and 30 physical servers intermittently fail with connection resets, while small pings and management access work. A large-packet ping from a backup client fails. Switch logs show one LACP member suspended, and the switchport for eth3 is in VLAN 1 rather than trunking VLAN 120. The backup clients are on a different subnet and use DD Boost over 10 GbE. Replication to the DR site traverses a WAN link that does not support jumbo frames. The customer wants to know whether MTU, VLAN tagging, or LACP is causing the failures and what to change.
1. Diagnose the likely causes of the intermittent DD Boost failures by ordering the checks you would perform on the DD and the switches.
Read the worked answer
Start at the DD with net show config and the interface/bond status to confirm bond0 is LACP, eth2/eth3 are members, VLAN 120 is tagged correctly, and MTU is 9000. Use net show hardware and net show stats to check link state, errors, drops, and whether one member is down. Test with a large-packet ping from a client on VLAN 120; if small packets pass and 9000-byte packets fail, there is an MTU mismatch or a device in the path that does not support jumbo frames.
On the switches, verify both member ports have the same LACP mode, speed, duplex, and VLAN configuration. A port left in VLAN 1 instead of trunking VLAN 120 explains partial connectivity and LACP suspension. Confirm the DD port is a trunk allowing VLAN 120 if the DD interface is tagged, or an access port in VLAN 120 if the DD interface is untagged; mixing these causes intermittent resets. Check switchport MTU and global jumbo-frame configuration. Because replication crosses a WAN that does not support jumbo frames, keep the replication interfaces at 1500 or ensure fragmentation/PMTUD works; do not force MTU 9000 across that path. The likely combination is one LACP member misconfigured, a VLAN mismatch on eth3, and an end-to-end MTU mismatch.
2. Recommend the corrected configuration for bond0, bond1, and the switchports, and explain how to validate it.
Read the worked answer
For bond0, configure eth2 and eth3 as an LACP bond carrying tagged VLAN 120, with MTU 9000 on the DD and on both switchports. Both switchports should be in the same LACP port-channel, trunking VLAN 120, with matching speed and MTU. Do not leave one port in VLAN 1. For bond1, use LACP with VLAN 220, but set MTU 1500 because the WAN does not carry jumbo frames; if the replication path requires 9000 on the LAN, terminate the large MTU at the LAN switch and let routing handle the WAN MTU. Keep management on ethMa separate from data bonds.
Validate in order: confirm LACP is up on both members, check VLAN membership and MAC learning, then test small and large pings from a client on VLAN 120. Test DD Boost with a small backup and a large backup, and monitor interface errors and drops. For replication, test with the actual WAN path and confirm no fragmentation or PMTUD black holes. Document the MTU per path rather than applying 9000 globally.
3. What would you tell the customer about using jumbo frames on the backup VLAN versus the replication WAN, and what ongoing monitoring is needed?
Read the worked answer
Jumbo frames are a LAN optimization, not a requirement. They reduce per-packet overhead on the 10 GbE backup VLAN when every device in the path supports them, including clients, switches, routers, and the DD. If any device is at 1500, large packets may be dropped and connections reset, which matches the intermittent backup failures. The safe approach is to standardize 9000 on the backup VLAN only after end-to-end validation, and keep the replication WAN at 1500 unless the carrier explicitly supports and tests 9000.
Monitor interface error counters, drops, LACP status, and DD Boost connection statistics after the change. Watch for resets during large transfers and correlate with switch logs. Also monitor replication throughput and latency separately, because the WAN path has different MTU and QoS characteristics. If jumbo frames cannot be guaranteed end-to-end, use 1500 on the backup VLAN; the performance loss is usually preferable to intermittent failures.