Packet Loss During VM vMotion Due to Delayed RARP Transmission

Products

• VMware vSphere ESXi

Issue/Introduction

Symptom

• Packet loss occurs during VM vMotion operations. In scenarios involving VMs with, for example, 10 virtual Network Interface Cards (vNICs), the packet loss can sometimes extend beyond 10 seconds when multiple VMs undergo vMotion simultaneously to same destination host.
• The VMs' vNICs are connected to a Virtual Distributed Switch (VDS) version 7.0, and the ESXi host is running version 7.

Environment

VMware vSphere ESXi 7

Cause

Following a vMotion, the destination ESXi host typically sends a Reverse Address Resolution Protocol (RARP) packet for the VM connected to it (usually within less than 1 second), as outlined in Knowledge Base (KB) article 343401. This allows the physical switch to update its records with the VM's new location.

However, in the specific combination of VDS 7.0 and ESXi 7, there is an issue causing a delay in the transmission of this RARP packet. The delay is particularly pronounced for VMs with multiple vNICs.

Resolution

To resolve this issue, upgrade the ESXi host to version 8.0.3 or a later release.

Additional Information

For general information on RARP, refer to the following link:
https://knowledge.broadcom.com/external/article/343401