MAC address of a vTEP interface is flapping on the fabric.

search cancel

book

calendar_today

VMware NSX

VM(s) connected to overlay segment.
Network I/O control reservations, or NUMA is configured on/for the VM(s).
Host Transport Node(s) configured with 3 or more vTEP interfaces.
TEP HA is configured for Host Transport Nodes.
MAC address of NSX TEP is observed flapping on different switchports on the fabric.
MAC address flaps may cause disconnects and packet loss.

When Network I/O Control, or NUMA is configured for the VM(s), the teaming policy set by NSX is overridden by the vDS.
Due to this override, TEP HA would not change the teaming at the vDS level, but would change internal teaming database.
This would result in vdl2 to use incorrect VTEP for encapsulation over an uplink for the VM traffic, leading to frequent move of vTEP's MAC address on the switch (TOR).

This is a known issue impacting VMware NSX.

You can use any one of the methods listed below to workaround the issue:

Disable VTEP HA feature and then confirm reproducibility of the issue. If the issue still persists, power the impacted VMs off and on to resolve the issue.
or
Disable Network I/O Control or NUMA reservations for VMs on overlay segments. If the issue still persists, power the impacted VMs off and on to resolve the issue.
or
Change uplink profile to use two vTEPs for overlay with TEP HA and Network I/O control enabled for the overlay VMs.

Note: If there are no overlay VMs with Network I/O Control, or NUMA configured on the host, then customer need not disable TEP HA.

thumb_up Yes

thumb_down No