We did some nic teaming failover/failback testing on a ESX host and pinged a VM running on this host during these events. We noticed some lost packets, surprisingly the failback lost clearly more packets than the failover. ----->>> failover 1 ping lost, failback 6 or 7 pings lost
We use Dell M630's in a M1000 chassis with two Dell MLX switches (vlt), these are connected to Alacatel switches with LACP. In vSphere we use a distributed switch with two 10G uplinks. Here are the settings for the switch.
Load Balancing: Route based on physical nic load
Network Failover Detection: Link Status Only
Notify Switches: Yes
Failback: Yes
Active Adapters: dvUplink1, dvUplink2
Standby Adapters: none
Unused Adapters: none
Is it normal behavior that a failback results in more dropped packets?
Is there a setting that I need to change to improve this?