Administrators may observe that the Maximum Transmission Unit (MTU) on a Distributed Virtual Switch (DVS) is configured to 1700. This typically occurs in environments utilizing VCF, NSX and VMware HCX, where HCX virtual machines (Interconnect/Network Extension appliances) reside on a port group associated with the DVS. Customers may question why this value is set above the standard 1500 bytes and seek the proper procedure to safely revert the configuration. Questions often arise regarding whether this MTU size is required for all cluster communication or only for hosts utilizing NSX overlay networking.
VMware Cloud Foundation 9.x
VMware NSX
VMware HCX
The DVS MTU of 1700 is a hard requirement for NSX Geneve encapsulation and HCX Network Extension (NE) operations. Tunneling protocols add encapsulation overhead (such as the Geneve header and HCX IPsec encryption) to standard Layer 2 traffic. An MTU of 1700 is the minimum recommended value to ensure that inner packets (standard 1500 MTU) can be encapsulated and transmitted across the underlay network without fragmentation or packet drops.
If the objective is to revert the DVS MTU to the standard 1500 bytes, the services relying on the increased buffer must be decommissioned first. Reverting the MTU while HCX VMs are active will cause immediate data plane failures for extended networks due to encapsulated packets exceeding the 1500-byte MTU limit.
To safely revert the MTU, perform the following steps:
NSX Manager UI.1500 via the vSphere Client.
Note:
The MTU 1700 requirement applies strictly to ESXi hosts within a cluster that utilize NSX overlay networking. In VCF 9 environments where NSX is not deployed, all datastores utilize NFS storage, and inter-site connectivity is strictly for management/inventory, an MTU of 1700 is not required for the inter-site path.
For more information regarding underlay MTU requirements and encapsulation overhead, refer to the following documentation: