VMware ESXi and vSphere Cluster Management

Create a Standard vSwitch and VMkernel Port Group for vMotion on VMware ESXi

Learn how to create a standard vSwitch and VMkernel port group with vMotion enabled on every ESXi host using the vSphere Web Client.

vSphere vMotion moves a powered-on virtual machine from one ESXi host to another with minimal interruption. For vMotion to work, the participating hosts need a compatible host-to-host network path.

vMotion traffic uses a dedicated VMkernel adapter, usually named vmkN. It does not use a virtual-machine port group. Configure a VMkernel adapter with the vMotion service enabled on every source and destination ESXi host that will participate in migrations.

How vMotion Networking Works

A typical vMotion path contains four related ESXi networking objects:

Physical NIC (vmnic) — A physical ESXi network adapter, such as vmnic2, that connects the host to the physical network.

Standard vSwitch — A virtual switch configured independently on one ESXi host. It connects virtual networking objects to assigned physical uplinks.

VMkernel port group — A port group that supplies the network label, VLAN settings, and network policy for host traffic.

VMkernel adapter (vmk) — The host networking interface that receives an IP address and provides services such as vMotion, management, storage, vSAN, or Fault Tolerance logging.

The uplink associates the standard vSwitch with a physical NIC. The VMkernel port group is created on that switch, and the VMkernel adapter is attached to the port group. Selecting the vMotion service on the adapter designates it for vMotion traffic.

Plan the vMotion Network Before Configuration

Hosts and vCenter

At least two ESXi hosts must be managed by the same vCenter environment for a normal vMotion migration. Every host that may be a source or destination needs a reachable VMkernel IP address on the planned vMotion network.

Connectivity design

The VMkernel adapters must have Layer 2 connectivity, or Layer 3 IP connectivity when the environment is designed to route vMotion traffic between subnets. In a routed design, provide the appropriate gateway configuration and verify the complete routed path before attempting migration.

A dedicated vMotion network or dedicated physical uplinks are recommended because they improve traffic isolation, predictability, and available capacity. In a small environment, vMotion may share a network with management traffic, but the shared design must have sufficient bandwidth and appropriate security and policy controls.

Addressing and VLAN planning

  • Assign a unique IP address to the vMotion VMkernel adapter on every host.
  • Record the subnet mask or prefix length for each address.
  • Provide a gateway when the vMotion design requires routed connectivity. A gateway is not normally needed for hosts communicating within the same local subnet.
  • Choose a VLAN ID when the vMotion network is VLAN-backed.
  • Ensure the physical switch ports and trunks carry the VLAN assigned to the VMkernel port group.
  • Do not assume that identical port-group names create connectivity. The VLAN, IP addressing, routing, physical uplink, and switch configuration must also match.

Bandwidth, MTU, and physical uplinks

Choose a physical NIC that is linked and connected to the intended vMotion network. Do not select an adapter that is disconnected, connected to the wrong physical-switch network, or reserved for another network purpose.

Plan sufficient bandwidth for the number and size of migrations. Check for contention with management, virtual-machine, storage, or other host traffic. The MTU must be consistent across the vMotion path. The standard MTU is 1500. An MTU larger than 1500 uses jumbo frames and requires end-to-end support on the VMkernel interface, vSwitch, physical switches, routers, and every other interface in the path. Do not enable jumbo frames only on ESXi.

Configuration worksheet

ESXi Host | Standard vSwitch | Physical Uplink | VMkernel Adapter | Port Group Label | VLAN ID | IP Address/Prefix | Gateway | MTU | vMotion Enabled

ESXi-01 | vSwitch-vMotion | vmnic2 | vmk1 | vMotion | 120 | 192.168.120.11/24 | Not required for local subnet | 1500 | Yes

ESXi-02 | vSwitch-vMotion | vmnic2 | vmk1 | vMotion | 120 | 192.168.120.12/24 | Not required for local subnet | 1500 | Yes

Create a VMkernel Port Group on a New Standard vSwitch

The following workflow uses the vSphere Web Client. Interface labels can vary slightly between vSphere versions, but the required choices are the same.

  1. Select the target ESXi host in the inventory.
  2. Open Manage, then open Networking.
  3. Start Add Networking.
  4. For the connection type, select VMkernel Network Adapter.
  5. When the workflow asks how to connect the adapter, select New standard switch if vMotion should use a new, dedicated standard switch. If an existing standard switch already has the correct uplinks and VLAN reachability, use that switch instead.
  6. Select an available physical network adapter, such as vmnic2, and assign it as the new switch's uplink. Confirm that the NIC has physical connectivity to the intended vMotion network.
  7. Give the VMkernel port group a meaningful network label, such as vMotion.
  8. Set the VLAN ID when the vMotion network uses VLAN segmentation. For example, enter 120 when the physical network provisions VLAN 120 for vMotion.
  9. On the service-selection step, enable the vMotion traffic service.
  10. Configure IPv4. Select DHCP only when your infrastructure intentionally provides DHCP for this host service. For normal infrastructure deployments, select manual or static addressing.
  11. For manual IPv4 configuration, enter the unique IP address, subnet mask or prefix length, and gateway when required by the routing design.
  12. Review the summary. Confirm the switch, uplink, port-group label, VLAN, MTU, IP configuration, and vMotion service selection.
  13. Finish the wizard.

DHCP versus static IPv4 configuration

DHCP automatically assigns IP configuration to the VMkernel interface. It can be appropriate in a controlled design, but address changes can complicate predictable host-to-host connectivity.

A static IP address is manually assigned and remains predictable. Static addresses are the normal choice for infrastructure networks such as vMotion. Each ESXi host must have a different address, and the addresses must belong to the intended vMotion network or be routable according to the design.

Repeat the configuration on every participating host

Repeat the workflow for every source and destination ESXi host. Use the same intended network design, including compatible VLAN, MTU, and uplink connectivity, while assigning host-specific VMkernel IP addresses. For example, ESXi-01 may use 192.168.120.11/24 and ESXi-02 may use 192.168.120.12/24.

Existing Standard vSwitch or New Standard vSwitch?

Creating a new standard vSwitch is useful when vMotion needs a dedicated uplink or a separately managed traffic path. However, creating duplicate switching infrastructure is unnecessary when an existing standard vSwitch already provides the correct physical uplinks, VLAN reachability, MTU, and capacity. In that case, add a VMkernel port group to the existing switch.

The key requirement is not the switch name. The key requirement is a complete, working path from the vMotion VMkernel adapter on one host to the vMotion VMkernel adapter on another host.

Example: Dedicated vMotion Network

In this example, two hosts use VLAN 120 for vMotion:

  • ESXi-01 creates a standard switch using vmnic2.
  • The switch has a VMkernel port group labeled vMotion.
  • The port group uses VLAN ID 120.
  • The vMotion VMkernel adapter uses 192.168.120.11/24 and has vMotion enabled.
  • ESXi-02 uses a corresponding uplink and VLAN 120 configuration.
  • ESXi-02 uses the different address 192.168.120.12/24 with vMotion enabled.
  • The physical switch configuration allows both uplinks to reach VLAN 120.

Both VMkernel adapters must be able to communicate before a live migration is attempted.

Example: Routed vMotion

In a routed design, participating hosts may use VMkernel addresses in different IP subnets. The network must route between those subnets, and each adapter receives the gateway configuration required by the design. Test connectivity between the VMkernel addresses before migration. A matching port-group label by itself does not provide routing.

Validate the Completed Configuration

  1. Inspect each standard vSwitch and verify that it has the expected physical uplink.
  2. Inspect the VMkernel adapter and verify that it exists, has the expected IP address, and has the vMotion service enabled.
  3. Compare the VLAN ID and MTU on every participating host.
  4. Confirm that physical switch ports provide the required VLAN and link connectivity.
  5. Use VMkernel connectivity testing between the hosts.
  6. Confirm that vCenter recognizes vMotion as available for the relevant hosts and virtual machine.
  7. Perform a controlled vMotion test and monitor the task result in vCenter.

Optional ESXi validation commands

Run these commands from an ESXi shell or through an approved remote administration method. Replace vmkN and the peer address with values from your environment.

esxcli network ip interface list
esxcli network ip interface tag get
esxcli network nic list
vmkping <peer-vmotion-vmkernel-ip>
vmkping -I vmkN <peer-vmotion-vmkernel-ip>
vmkping -I vmkN -d -s 8972 <peer-vmotion-vmkernel-ip>

The final command is an example jumbo-frame test for an MTU 9000 design. Use it only after the complete path has been configured for jumbo frames. The -I vmkN option sources the test from the selected VMkernel adapter. The -d option prevents fragmentation, and -s 8972 tests a large payload appropriate for an MTU 9000 path.

Check | Expected Result | Likely Issue If It Fails

Physical uplink | The expected vmnic is assigned, linked, and connected to the correct switch network | Disconnected NIC, wrong cabling, disabled switch port, or wrong uplink

VMkernel adapter | The adapter exists, has a unique expected IP address, and has vMotion enabled | Wrong service selection, duplicate address, or incorrect adapter

VLAN | The port group and physical switch carry the intended VLAN | VLAN mismatch, incorrect access/trunk configuration, or missing VLAN provisioning

Basic vmkping | The peer vMotion VMkernel address responds | Incorrect subnet, gateway, route, firewall, uplink, or physical-switch configuration

Jumbo vmkping | A non-fragmented large-packet test succeeds | Inconsistent MTU somewhere along the path

vCenter migration readiness | vCenter offers vMotion for the selected hosts and virtual machine | Host compatibility, network, storage, CPU, licensing, or other migration prerequisite

Troubleshoot Common Problems

Hosts cannot communicate over the migration network

  • Confirm that every participating host has a VMkernel adapter with vMotion enabled.
  • Check each VMkernel IP address, subnet mask or prefix, and gateway configuration.
  • Run vmkping to the peer address, explicitly sourcing the intended vMotion adapter.
  • Check that the selected physical NIC is linked and connected to the correct physical-switch network.

VMkernel adapters cannot reach each other on the intended VLAN

  • Compare the port-group VLAN ID on every host.
  • Verify that the physical switch ports carry the required VLAN.
  • Check whether the switch ports use the access or trunk behavior appropriate for the ESXi uplink design.
  • Ensure that the VMkernel adapters do not have overlapping or duplicate IP addresses.

vMotion is intermittent or slow

  • Check bandwidth contention when vMotion shares uplinks with management, virtual-machine, storage, or other host traffic.
  • Review uplink speed, duplex, error counters, and physical-switch health.
  • Confirm the traffic-isolation and capacity design.
  • Check whether one host uses a different MTU or a different physical path.

Jumbo-frame testing fails

  • Verify the VMkernel adapter MTU on all hosts.
  • Verify the standard vSwitch MTU where applicable.
  • Verify physical-switch and routed-interface MTUs across the entire path.
  • Test standard-sized packets first, then run a non-fragmented large-packet vmkping test.

vMotion is enabled on the wrong VMkernel adapter

  • Identify the adapter that has the vMotion service tag.
  • Confirm that it uses the intended port group, VLAN, IP subnet, and physical uplink path.
  • Move or recreate the service configuration on the correct adapter if necessary.

Exam-Relevant Notes

  • vMotion uses a VMkernel adapter, not a virtual-machine port group.
  • Every participating ESXi host needs a unique, reachable vMotion VMkernel IP address.
  • A standard vSwitch is local to an ESXi host; configure compatible networking on each host.
  • The VMkernel port group carries the label, VLAN, and service settings for the adapter.
  • Identical port-group names do not guarantee connectivity.
  • Jumbo frames require end-to-end MTU consistency, not just an ESXi-side setting.
  • A dedicated network is recommended, but a shared management and vMotion design can be used in small environments when capacity and policy permit.

After the network is validated, confirm the migration prerequisites in vCenter and complete a controlled test migration. Related host communication concepts are covered in Communication Between vCenter Server and ESXi. For broader study, see the VMware ESXi Online Course.