Skip to content

How to Configure NIC Teaming on Windows Server Print

  • 0

Applies to: Windows Server 2019, Windows Server 2022 and Windows Server 2025, subject to the workload and edition requirements below.

Choose LBFO or Switch Embedded Teaming deliberately

NIC teaming combines supported adapters to provide traffic distribution and resilience against selected link or adapter failures. Traditional Windows NIC Teaming is often called LBFO, for load balancing and failover. Switch Embedded Teaming, or SET, integrates teaming into the Hyper-V virtual switch. They are different configurations with different support boundaries; a command that creates an LBFO team is not a modern Hyper-V SET deployment.

Windows Server 2019 can have older Hyper-V designs that use LBFO. Starting with Windows Server 2022, Microsoft directs Hyper-V virtual switches to SET rather than binding them to LBFO teams. LBFO remains supported for other non-Hyper-V scenarios. This guide shows a non-Hyper-V LBFO example and a separate SET creation pattern. Choose one that matches the actual role; do not build both on the same adapters.

Design for the failure you intend to survive

A two-adapter team does not automatically survive a switch failure, a shared cable path or a provider outage. Identify adapter, switch, power and upstream failure domains. Switch-independent teaming can use separate switches where the network supports the design. LACP and static switch-dependent modes require compatible configuration on the upstream network and may depend on a supported switch stack or multi-chassis arrangement.

Teaming does not make every single connection run at the sum of all link speeds. Traffic distribution depends on the selected algorithm and flow characteristics. A single large transfer can remain limited to one member’s capacity. Choose throughput expectations from the actual workload and test multiple representative flows. Do not promise that two 10 Gb adapters provide a 20 Gb pipe for every application.

Review specialized traffic such as RDMA/SMB Direct, iSCSI, clustering and storage networks before teaming. Traditional LBFO and SET have different offload behaviour. Supported MPIO designs are usually the relevant path for block-storage redundancy rather than an improvised team. In a VPS, the underlying hypervisor and provider must support the proposed presentation; two guest adapters do not establish two independent physical links.

Inventory adapters and preserve the network baseline

Use console or out-of-band access and schedule the change. Record IP addresses, DNS, routes, VLANs and physical port mappings. Confirm supported adapter models, driver and firmware versions, link speeds and settings. SET has stricter homogeneity requirements than an arbitrary collection of network cards; use a compatible, deliberately matched set.

Get-NetAdapter | Select-Object Name,InterfaceDescription,ifIndex,MacAddress,Status,LinkSpeed
Get-NetIPConfiguration
Get-NetRoute -AddressFamily IPv4
Get-NetLbfoTeam
Get-NetLbfoTeamMember
Get-VMSwitch -ErrorAction SilentlyContinue

Verify that each proposed member is unused by another team or virtual switch and does not carry an unaccounted production dependency. Do not select all Up adapters as team members. Name the two reviewed uplinks explicitly and keep an independent recovery path. Team creation can reconfigure the interfaces and interrupt the very remote session used to perform it.

Create a non-Hyper-V LBFO team

For a supported ordinary server network, the example creates an LBFO team with switch-independent mode and Dynamic load balancing. The upstream ports must support the agreed VLAN configuration, but this mode does not require an LACP bundle. Replace NIC01 and NIC02 with the verified adapters and do not use this pattern as a workaround for a newer Hyper-V switch requirement.

New-NetLbfoTeam -Name 'ServerTeam' -TeamMembers 'NIC01','NIC02' `
  -TeamingMode SwitchIndependent -LoadBalancingAlgorithm Dynamic
Get-NetLbfoTeam -Name 'ServerTeam'
Get-NetLbfoTeamMember -Team 'ServerTeam'
Get-NetAdapter -Name 'ServerTeam'

Assign or verify the approved host IP, DNS and routes on the resulting team interface. Do not leave conflicting static addresses on the member adapters or add duplicate default gateways. If the design uses a tagged team interface, configure its VLAN through the supported team-interface settings and verify the upstream ports. A link being Up does not prove a VLAN is correct.

Use SET for a new Hyper-V teaming design

On a Hyper-V host that requires a teamed external switch, create the virtual switch with embedded teaming on the explicitly selected uplinks. SET uses switch-independent teaming; do not configure its upstream ports as an ordinary LACP team merely because two links exist. Confirm the current driver and hardware requirements before creation.

# Alternative design: run on the approved Hyper-V host, not on the LBFO members above.
New-VMSwitch -Name 'Converged-SET' -NetAdapterName 'HV-Uplink01','HV-Uplink02' `
  -EnableEmbeddedTeaming $true -AllowManagementOS $true
Get-VMSwitchTeam -Name 'Converged-SET'
Get-VMNetworkAdapter -ManagementOS
Get-NetIPConfiguration

The host’s management address belongs on the appropriate management vEthernet adapter, with its own VLAN and routing policy. Guest adapters also need deliberate VLAN and bandwidth configuration. This command creates a team-backed switch, not a complete converged network with storage, migration and quality-of-service policy. Do not attach an LBFO team as the uplink to bypass the supported design on Windows Server 2022 or 2025.

Verify connectivity and member health

Check team state, individual member state and the resulting logical interface. Verify that the intended host and guest addresses use the new interface and that the default and specific routes remain correct. Test DNS, management access and actual application ports from another authorized machine. Locally originated tests alone can miss an inbound path problem.

# LBFO example:
Get-NetLbfoTeam -Name 'ServerTeam' | Format-List *
Get-NetLbfoTeamMember -Team 'ServerTeam'
Get-NetAdapterStatistics -Name 'NIC01','NIC02'
# SET example, when that alternative was selected:
Get-VMSwitchTeam -Name 'Converged-SET'

During a controlled maintenance test, fail one intended link at a time through a supported method while monitoring application continuity and management access. Keep the other path verified and have a console recovery option. Restore the link and confirm the team recovers. Do not perform cable-pull experiments on production storage traffic without the workload owner’s plan and an understood outage consequence.

Diagnose distribution and failover problems

If one member carries most traffic, compare the traffic pattern with the chosen algorithm before calling it a fault. Single-flow limitations, inbound switch decisions and active/standby configuration can explain the pattern. Test several independent flows and inspect errors, discards and switch counters. A throughput graph without workload context does not establish a teaming defect.

If failover loses connectivity, inspect VLAN consistency, switch forwarding, MAC learning, port security, provider anti-spoofing and whether the remaining path actually reaches the same network. If the team will not create, check member ownership, adapter support and driver state. Do not disable offloads or force hidden compatibility switches broadly before identifying the failing component. A teaming change should not become a collection of undocumented performance tweaks.

Rollback to a known reachable interface

Before removing a team or SET switch, inventory all host and VM dependencies and preserve its IP, VLAN and route settings. A logical interface can disappear with its addresses when the object is removed. Restore the original approved configuration to the intended standalone adapter from the console, then verify remote management and application traffic before ending the maintenance window.

For an established host, migrate connected VM adapters and specialized traffic through a reviewed plan; removing the switch affects more than the host’s RDP connection. Keep the final configuration, member serial or MAC identities, switch-port mappings, mode, algorithm and failure-test evidence documented. Recheck after driver, firmware and Windows Server upgrades. A team provides useful resilience only when the remaining path has been demonstrated to carry the actual workload under the failures the business expects it to tolerate.

Related guidance and official references

Create an external Hyper-V switch; Manage static routes; Diagnose network connectivity.


Was this answer helpful?

« Back