During a planned network cutover (e.g., VLAN to VXLAN transitioning), ESXi hosts may become disconnected from the vCenter Server and show as "Not Responding" in the vSphere Client inventory. Virtual machine workloads continue to run without interruption, but vCenter management operations are impaired. This symptom presents when the default gateway routing changes.
VMware ESXi 8.0u3
VMware vCenter Server 8.0u3
The Default Gateway for the Management Network on the host is changed to another piece of hardware that contains a different MAC address
The ESXi host ARP cache may retain the stale/old physical MAC address of the old default gateway because a Gratuitous ARP (GARP) was not received or processed during the physical network cutover. Outbound management traffic (Port 902) routing fails until the ESXi ARP timer expires.
Prior to the network cutover, temporarily disable vSphere HA and DRS on the affected cluster to prevent host isolation responses.
Ensure Out-of-Band (OOB) management access or ESXi Shell access is available from an adjacent subnet jumpbox.
If hosts enter a "Not Responding" state and do not recover automatically post-cutover, log into the ESXi Shell.
Validate the current ARP neighbor entries by executing: esxcli network ip neighbor list
Manually remove the stale MAC address entry for the default gateway: esxcli network ip neighbor remove -v <VMK_INTERFACE> -a <GATEWAY_IP>
Force an immediate ARP request to the new gateway MAC address: vmkping -I <VMK_INTERFACE> <GATEWAY_IP>
If the gateway IP address itself has changed and needs manual updating, update the gateway via the Direct Console User Interface (DCUI) or via ESXCLI: esxcli network ip interface ipv4 set --interface-name <VMK_INTERFACE> --gateway <NEW_GATEWAY_IP> --type static --ipv4 <HOST_IP> --netmask <SUBNET_MASK>
Verify that the vCenter Server regains connectivity to the affected ESXi hosts. Re-enable vSphere HA and DRS once the management network is stable.