VMs on NSX segments lose network connectivity when ESXi host uplinks are configured with EtherChannel on the physical switch

Products

• VMware NSX

Issue/Introduction

• VM configured on NSX Segment (Vlan or Overlay) is unreachable over the network with the corresponding uplinks configured with EtherChannel on physical switch ports.
• Same VMs when configured on VDS switch port group that is enabled with IP Hash load balancing policy is reachable.

Environment

VMware NSX

Cause

• If EtherChannel on the Physical Switch is used, TX and RX uplinks are projected separately for vNICs of the VM that is configured on a segment/portgroup overlay or vlan (without IP Hashing policy) and therefore the traffic will not work.
• Hence, there is a high possibility that traffic can go via uplink vmnicX and the responses can be seen on uplink vmnicY.

Resolution

• NSX segments (both VLAN and Overlay) does not support IP Hash Load Balancing Policy
• Disable the EtherChannel on physical ports of the switch where the uplinks are connected.

 

Additional Information

• Alternatively, to implement port-link aggregation LAG in NSX refer to the documentation Configure an ESXi Host Transport Node with Link Aggregation Group
• Supported LAG hashing mechanism options for LACP link aggregation can be referenced here Possible LAG hashing mechanism options