guest VM connectivity loss due to hyper-v SET dynamic load balancing and upstream switch MAC learning limits

Question

guest VM connectivity loss due to hyper-v SET dynamic load balancing and upstream switch MAC learning limits

Rossia Luca 0

We've obsserved this behavor:

Successfully initializes synthetic network adapters across all guest operating systems.

Establishes physical uplink connectivity across all teamed network interface cards.

Distributes outbound frame transmissions across aggregated host physical ports.

Resolves initial address resolution protocol requests across the virtual switch fabric.

Microsoft-Windows-Hyper-V-VmSwitch Event ID 113 logs team member operational status as healthy with zero hardware drop counters.

TCP handshake state synchronization completes reliably during initial connection establishment.

Virtual switch port security profiles maintain consistent tag encapsulation across active VLANs.

but return ingress traffic never reaches the guest virtual network adapters (upstream switch MAC address tables reach maximum capacity and target port forwarding registers remain empty)?

Is there a required Windows Server update, or host-side configuration fix for Hyper-V SET load balancing algorithm enforcement to lock down HyperVPort mode over Dynamic?

0 comments

1 answer

Your answer

Answer 1

Hello @Rossia Luca

Based on the symptoms, I would focus on the upstream switch MAC-learning behavior rather than Event ID 113 or a Windows update. If outbound traffic works but return traffic disappears once the switch MAC table reaches its limit, that strongly suggests the physical switching layer is no longer learning or forwarding the VM MAC addresses correctly.

For SET, Windows supports both Dynamic and HyperVPort load-balancing algorithms, and you can explicitly change an existing SET switch to Hyper-V Port mode:

Get-VMSwitchTeam -Name "YourSETvSwitch"
Set-VMSwitchTeam -Name "YourSETvSwitch" `
    -LoadBalancingAlgorithm HyperVPort

Microsoft's current Hyper-V cluster networking guidance recommends Switch Independent teaming with Hyper-V Port load balancing for clustered Hyper-V deployments. Hyper-V Port associates a VM's virtual NIC/MAC with a particular physical team member, which can also make MAC learning more predictable for the upstream switches.

I would therefore test HyperVPort and simultaneously check the physical switches for:

MAC-address-table capacity or per-port MAC limits.
Port-security/maximum learned MAC settings.
MAC flapping between SET uplinks.
Whether the affected VM MAC is learned on the expected physical port.
VLAN/trunk configuration consistency across every SET member.

If connectivity becomes stable immediately after switching from Dynamic to HyperVPort, that would strongly point toward an interaction with upstream MAC learning rather than a guest or vSwitch failure.

I wouldn't assume a Windows hotfix is required based on the information provided. The fact that the upstream MAC table is reaching its configured limit needs to be addressed as well; changing the SET algorithm shouldn't be treated as a substitute for correcting an undersized or restrictive MAC-learning configuration.

Please "Accept the Answer" if this information helped you. This will help us and others in the community as well.