VMware ESXi and vSphere Cluster Management

Configure Network Failure Handling and NIC Failover in VMware vSphere

Learn how to configure Notify switches, failback, and Active, Standby, and Unused uplinks on a VMware ESXi standard vSwitch.

Network failure handling in VMware vSphere determines how an ESXi host keeps virtual machines connected when a physical network path becomes unavailable. By combining NIC teaming with a failover policy, a virtual switch can move traffic from a failed uplink to another eligible physical adapter.

This lesson focuses on standard vSwitch networking configured through the vSphere Client. It covers switch notifications, failback, uplink roles, failure detection, configuration, validation, and troubleshooting.

For a broader failure-handling overview, see How To Handle Network Failures.

How vSphere network failover works

An ESXi host connects virtual machines to external networks through several layers:

Virtual NIC: The network interface presented to a virtual machine.
Virtual switch: The software switch that connects virtual NICs to port groups and physical uplinks.
Physical NIC uplink: A host adapter, commonly displayed as a vmnic, that connects the virtual switch to the physical network.
Physical switching infrastructure: The upstream switches, links, VLANs, trunks, and routing paths beyond the ESXi host.

A failover policy primarily controls how the virtual switch uses its physical uplinks. It does not repair a failed physical switch, correct a missing VLAN, or guarantee that every upstream path is redundant. The backup uplink must lead to a usable network path.

Teaming means using multiple physical adapters for redundancy and, depending on the selected load-balancing policy, traffic distribution. Failover is the movement of traffic from an unavailable uplink to another eligible uplink.

These policies are configured at the scope selected in the vSphere Client. Depending on the networking object and editor being used, the settings may apply to a virtual switch or to a port group that overrides the switch-level policy.

Teaming and failover policies

Policy: Notify switches

What it controls: Whether the virtual switch sends network notifications when a virtual machine network interface connects or moves to another physical uplink after failover.

Preferred behavior: Enabled in normal designs.

Operational consideration: The notification helps upstream switches refresh MAC address forwarding information. A specialized network design may require a different setting.

Policy: Failback

What it controls: What happens when an uplink that previously failed becomes healthy again.

Preferred behavior: Enable it when traffic should return automatically to the preferred active uplink; disable it when avoiding an immediate second traffic transition is more important.

Operational consideration: Recovery behavior must match the intended traffic placement and maintenance plan.

Policy: Failover order

What it controls: Which physical adapters may carry traffic and the priority assigned to them.

Preferred behavior: Put normal-use adapters in Active, backup adapters in Standby, and deliberately excluded adapters in Unused.

Operational consideration: Every path expected to provide redundancy must have equivalent physical switch, VLAN, and upstream access.

Notify switches

Notify switches controls whether the virtual switch informs the physical network about a relevant virtual NIC connection or uplink movement. The notification is useful after failover because the virtual machine's traffic may now leave the ESXi host through a different physical switch port.

Physical switches maintain a MAC address table. This table maps learned device MAC addresses to physical switch ports. When a virtual machine moves from one host uplink to another, an upstream switch may still have the VM's MAC address associated with the old port.

With switch notifications enabled, the network receives information associated with the changed location and can update forwarding behavior promptly. If notifications are disabled, unavailable, or not processed as expected, traffic may continue toward the previous port until the physical switch relearns the MAC location.

The normal recommended state is enabled, unless a specific network architecture or operational requirement calls for another setting. Enabling the policy does not replace correct physical switch configuration.

Failback

Failback is the action taken after an uplink that previously failed becomes healthy again.

Failback enabled: A recovered preferred active uplink resumes service automatically. The temporary failover uplink returns to its standby role when the policy permits.
Failback disabled: The currently functioning failover uplink remains in use after the preferred uplink recovers. Traffic changes again only when another failure or policy event causes a transition.

Enabled failback restores the intended traffic placement and preferred uplink order. However, recovery itself causes another traffic transition. Disabled failback avoids that immediate transition, which can be useful when a recovering link may be unstable or when operational stability is preferred over restoring the original path immediately.

Failover order and uplink roles

Failover order establishes which physical adapters can carry traffic and their priority. The three roles are:

Active: An uplink normally eligible to carry traffic while connectivity is healthy. Multiple active uplinks may distribute traffic according to the selected load-balancing policy.

Standby: An uplink reserved for use when an active uplink is declared unavailable. It provides backup capacity rather than normal traffic service.

Unused: An adapter deliberately excluded from carrying traffic for the selected switch or port group. It may still be assigned elsewhere on the host.

For example, a production port group might assign vmnic0 as Active and vmnic1 as Standby. If vmnic0 fails, traffic can move to vmnic1. If failback is enabled, traffic returns to vmnic0 after recovery. If failback is disabled, traffic remains on vmnic1 until another event causes a change.

Multiple Active uplinks can provide load distribution, but the exact distribution depends on the selected teaming and load-balancing policy. Standby uplinks provide backup capacity and are not normally used for the regular traffic load.

Preferred active ordering and failback are related. When failback is enabled, a recovered uplink that has preference can resume service. When failback is disabled, the currently operating failover uplink remains in use even though the preferred uplink is healthy again.

Network failure detection context

Failover depends on how the virtual switch determines that an uplink has failed. Detection timing and reliability are influenced by the configured detection method, uplink topology, and physical-switch behavior.

Link-status detection

Link-status detection responds to the physical NIC reporting loss of its direct link. It is appropriate for failures such as an unplugged cable, a powered-off switch port, or a physical interface that reports link down.

It may not detect every failure beyond the directly connected link. For example, the local NIC and switch port can remain electrically up while an upstream path is broken.

Beacon probing

Beacon probing uses probe traffic to help identify certain upstream path failures that leave the local physical link up. It can provide additional visibility into path health, but its usefulness depends on the topology and the way uplinks connect to the physical network.

Choose and evaluate the detection method as part of the complete network design. A standby uplink cannot provide useful failover if it is connected to an unavailable switch, lacks the required VLAN, or reaches the same failed upstream path.

Configure a standard vSwitch in the vSphere Client

Select the target ESXi host in the vSphere Client.
Open the host's Manage area and choose Networking.
Select the relevant standard virtual switch.
Open the switch's edit or settings action.
Open the Teaming and failover section.
Set Notify switches according to the network design. Leave it enabled for the normal recommended configuration.
Set Failback to enabled if a recovered preferred uplink should resume service automatically. Set it to disabled if the currently functioning failover path should remain in use.
Arrange physical adapters into Active, Standby, and Unused categories.
Apply or save the changes.
Confirm that the intended uplinks are assigned to the switch, connected, and available.

Apply the policy at the correct scope. If a port group has its own teaming and failover override, changing the switch-level setting may not change that port group's effective behavior.

Practical configurations

One active and one standby uplink

Assign vmnic0 to Active and vmnic1 to Standby for a production VM port group. If vmnic0 or its usable path fails, the virtual switch moves traffic to vmnic1. With failback enabled, traffic returns to vmnic0 after recovery. With failback disabled, traffic stays on vmnic1.

Keeping traffic on the recovered failover path

Use a preferred active uplink and a standby uplink, but disable failback. After the preferred uplink recovers, the working failover uplink remains in use. This reduces an additional transition immediately after recovery.

Keeping physical switches aware of uplink movement

Leave Notify switches enabled for a port group with redundant host uplinks. When traffic moves to another uplink, the upstream switching environment can refresh its forwarding information more promptly.

Excluding an adapter from one network

Place an uplink in Unused for a selected port group when that adapter must not carry traffic for that network. The adapter can remain assigned to another network function elsewhere on the host.

Validation and operational planning

Map every assigned vmnic to the expected physical switch and switch port.
Confirm that active and standby uplinks have equivalent access to the required VLANs and upstream networks.
Verify that physical switch ports use the intended access or trunk configuration and are consistent with the ESXi design.
Use at least two suitably connected uplinks when redundancy is required.
Plan a controlled uplink failure test during an approved maintenance window.
During the test, verify that virtual machine traffic remains reachable through the backup path.
Restore the failed uplink and verify behavior against the configured failback setting.
Review physical-switch MAC learning and port status if forwarding does not converge correctly.

Failure and recovery behavior

Event: Preferred active uplink fails. With failback enabled, an eligible standby uplink takes over. With failback disabled, the standby uplink also takes over. The expected state is that the backup uplink carries traffic.

Event: Original uplink recovers. With failback enabled, traffic preference returns to the recovered preferred uplink and the temporary standby path returns to standby. With failback disabled, the currently functioning failover uplink remains in use.

Troubleshooting network failover

Virtual machines lose connectivity longer than expected

Possible causes include disabled switch notifications, an upstream switch that has not updated its MAC forwarding entry, or a backup uplink without equivalent VLAN or trunk access.

Confirm that Notify switches is enabled.
Check physical-switch MAC address learning and port state.
Compare VLAN, trunk, and port configuration for active and standby uplinks.

Traffic returns to the original NIC unexpectedly

If traffic moves back immediately after the preferred uplink recovers, failback is probably enabled.

Review the Teaming and failover settings.
Disable failback when traffic should remain on the currently operating failover path after recovery.

A standby adapter does not take over

The local physical link may still be up even though an upstream path has failed. Other causes include a detection method that does not match the topology or a standby adapter without equivalent network access.

Verify the selected failure detection method.
Inspect physical switch and upstream connectivity.
Validate VLAN reachability through the standby uplink.
Confirm that the adapter is actually assigned to Standby rather than Unused.

An expected redundant adapter never carries traffic

Check whether the adapter is assigned to Unused, is not attached to the target standard vSwitch, or has no usable physical link.

Review the Active, Standby, and Unused assignments.
Confirm that the vmnic is associated with the intended standard vSwitch.
Check adapter status and the corresponding physical switch port.

Exam-relevant notes

Notify switches helps upstream physical switches refresh MAC forwarding information after virtual NIC or uplink movement.
Failback enabled returns traffic preference to a recovered preferred uplink.
Failback disabled keeps traffic on the currently functioning failover uplink after recovery.
Active means normally eligible to carry traffic.
Standby means reserved for failover.
Unused means excluded from traffic for the applicable switch or port group.
Link-status detection primarily sees direct physical link loss; beacon probing can help identify some upstream failures while the local link remains up.
Failover settings cannot compensate for missing VLAN access, incorrect trunk configuration, or a failed shared upstream path.