ESXi Host Migration to New DVS with LACP Fails and Rolls Back

Products

• VMware vSphere ESXi
• VMware vCenter Server

Issue/Introduction

• Attempting to migrate an ESXi host's physical uplinks and management interface (vmk0) from a source Distributed Virtual Switch (DVS) to a new destination DVS with an identical Link Aggregation Group (LAG) configuration fails.
  
  
• The ESXi host temporarily loses network connectivity, and the vCenter Server triggers a network rollback, reverting the host back to the source DVS.
  
  
• The /var/log/lacp.log file on the affected ESXi host displays a sequence indicating port blockage and PDU failure:
  
  

  <TIMESTAMP> No(29) lacp[2098216]: 3602, Receive uplink vmnic# event (port blocked), smVars 0x3
  <TIMESTAMP> Er(27) lacp[2098216]: 2329, Failed to send lacpdu to uplink vmnic#
  <TIMESTAMP> Al(25) lacp[2098216]: 777, AddPort port(vmnic#/###_LAG_01-0) flag 0 systemMac <##:##:##:##:##:##> uplinkMac <##:##:##:##:##:##>
  <TIMESTAMP> Al(25) lacp[2098216]: 777, UWSetPortInfo port(vmnic#/###_LAG_01-0) flag 0 systemMac <##> uplinkMac <##>
  <TIMESTAMP> No(29) lacp[2098216]: 2158, Add uplink vmnic# into its own native aggregator 35, numPorts 1, agg <##>, portKey 0xf full 1
  <TIMESTAMP> No(29) lacp[2098216]: 2002, Detach uplink vmnic# from aggregator 35, numPorts 0, agg <##>
  <TIMESTAMP> No(29) lacp[2098216]: 2158, Add uplink vmnic# into its own native aggregator 35, numPorts 1, agg <##>, portKey 0xf full 1
  <TIMESTAMP> No(29) lacp[2098216]: 2002, Detach uplink vmnic# from aggregator 35, numPorts 0, agg <##>

Environment

VMware vSphere ESXi

VMware vCenter Server 

Cause

The migration fails at the physical network layer due to IEEE 802.3ad loop prevention mechanisms on the upstream physical switch. Every DVS generates a unique LACP System MAC address. When the physical uplinks are forcefully migrated to the new DVS, the ESXi host begins transmitting LACP Protocol Data Units (PDUs) using the new System MAC.

The upstream physical switch detects this abrupt MAC transition on the existing Port-Channel, identifies it as a split-brain anomaly or miswiring event, and actively places the physical ports into an err-disable (blocked) state. This drops all LACP negotiation and management traffic, forcing the vCenter network rollback safeguard to execute.

Resolution

• Evacuate all virtual machines from the ESXi host by placing it into Maintenance Mode.

• Coordinate with the network administration team to remove one physical port from the physical switch Port-Channel, configuring it as a standard 802.1Q trunk.

• Remove the corresponding physical uplink (vmnic) from the source DVS LAG and assign it to the destination virtual switch as a standard standalone uplink.

• Migrate the management interface (vmk0) to the destination virtual switch using the standalone uplink path.

• Once management connectivity is validated on the new virtual switch, dissolve the remainder of the Port-Channel on the physical switch.

• Migrate the final physical uplink (vmnic) to the destination virtual switch.

• Reconfigure the target physical switch ports and the destination virtual switch into a new LAG if LACP is required at the destination.

Additional Information

Attempting to migrate ESXi host management interface to a new Distributed Virtual Switch fails and rolls back

One uplink in a LAG is removed unexpectedly and falls back to a standalone link