ESXi host disconnections and VM migration failures due to Management Network MTU mismatch

Products

• VMware vSphere ESXi

Issue/Introduction

The primary product affected is VMware vSphere ESXi within the Networking component.

• ESXi hosts intermittently disconnect and rapidly reconnect to the vCenter Server management console.
• Virtual Machine migrations (including VMs with Raw Device Mapping - RDM disks) fail between standalone hosts.
• Network diagnostic tests (vmkping) show 100% packet loss when using large payloads, despite the hosts being reachable via standard ICMP pings.
• Incremental packet drops are observed on physical uplinks (e.g., vmnic5).

Environment

• VMware ESXi 8.x

Cause

The issues are caused by a network-layer MTU configuration mismatch on the Management Network (vmk0).

In this scenario, one host is configured with a standard MTU of 1500, while the peer host is configured for Jumbo Frames (9000). If the underlying physical network infrastructure does not consistently support Jumbo Frames across the entire path, packets exceeding 1500 bytes are dropped, leading to heartbeat loss and migration failures.

Resolution

To resolve the configuration mismatch and stabilize host connectivity, align the MTU settings across all ESXi hosts and the physical network.

Prerequisites

• Identify the current MTU settings for all affected hosts.
• Ensure you have administrative access to the vSphere Client.

Steps to Align MTU to Standard (1500)

1. Log in to the vSphere Client.
2. Navigate to the affected ESXi host in the inventory.
3. Select the Configure tab.
4. Under Networking, select Virtual switches.
5. Locate the switch associated with the Management Network (typically vSwitch0).
6. Click Edit settings for the switch.
7. Modify the MTU value to 1500.
8. Click OK to save the changes.
9. Repeat these steps for any other hosts configured with inconsistent MTU values.

Verification via Command Line

Perform a size-specific ping from the ESXi shell to verify the path:

• To test standard MTU (expected to pass):
  

  vmkping -I vmk0 -d -s 1472 <destination_IP>

• To test Jumbo Frames (will fail if mismatch exists):
  

  vmkping -I vmk0 -d -s 8972 <destination_IP>

Alternative Solutions
If Jumbo Frames are required for the environment but the physical network is not yet fully configured:

1. Physical Network Update: Coordinate with the networking team to ensure MTU 9000 is enabled end-to-end on all physical switches and routers.
2. Cold Migration: Perform a cold migration (powered off) of the VMs, or create a temporary Provisioning vmkernel interface on the destination host with an MTU of 1500 to facilitate the transfer.

Additional Information

Related KB: KB 318647 ESXi host disconnects intermittently from vCenter Server
Related KB: KB 313061 Testing jumbo frame pings from ESXi over the network

To speak with a customer representative or a Support Engineer, see Contact Support. Scroll to the bottom of the page and click on your respective region.