TEP tunnels down and NSX Manager UI unresponsive due to MTU mismatch in Azure VMware Solution (AVS).
Article ID: 409964
Updated On:
Products
Issue/Introduction
TEP (Tunnel Endpoint) tunnels between multiple ESXi transport nodes are in a DOWN state.
The Host Controller connectivity status for the affected ESXi hosts shows as "Disconnected."
The NSX Manager User Interface (UI) is extremely slow to load. Specific pages, such as the Node View Details, fail to render and remain in a perpetual loading/spinning state.
Packet captures taken during the event reveal that Bidirectional Forwarding Detection (BFD) down packets sent by the source host reach the destination uplink, but are not received by the destination VMkernel port.
The NSX Managers are deployed on Azure VMware Solution (AVS) infrastructure.
Environment
Cause
This issue is caused by a Maximum Transmission Unit (MTU) mismatch within the intermediate physical network path.
By default, NSX Manager nodes send control plane network traffic with the Don't Fragment (DF) bit set. If an intermediate network device between the NSX Manager and the ESXi hosts is configured with an MTU smaller than the default 1500 bytes, it cannot forward the packets and is forced to drop them.
This packet loss disrupts the critical control plane synchronization between the NSX Central Control Plane (CCP) and the host's Local Control Plane (LCP), resulting in dropped TEP tunnels, host disconnections, and NSX Manager UI performance degradation.
Resolution
To permanently resolve this issue, the MTU of the NSX Manager nodes must be lowered to accommodate the restrictive path MTU.
Adjust the MTU on the NSX Manager nodes from the default 1500 to 1400.
Lowering the MTU ensures that the control plane packets are small enough to successfully traverse all intermediate network devices without requiring fragmentation, even when the DF bit is set.
Once the MTU is adjusted, control plane communication will synchronize properly, restoring TEP tunnel stability and NSX Manager UI performance.
Feedback
