Network Connectivity Drop post Virtual Machine vMotion Due to Physical Switch ARP Delays
Article ID: 446371
Updated On:
Products
Issue/Introduction
Post the migration of a Virtual Machine (VM) via vMotion, a network connectivity drop lasting 5-10 seconds occurs. Simultaneous network packet captures from both the source and destination ESXi hosts reveal the following:
Destination Host: Upon completing the migration, the destination ESXi host successfully transmitted a Reverse ARP (RARP) broadcast. This is the standard mechanism used to update the physical switch's MAC address table with the VM's new location. We can see the VM successfully transmitting outbound traffic; however, the capture logs "No response found" for the incoming gateway replies.
Source Host: Even after the VM migrated to the new host and the RARP was sent, the reply packets from the 192.168.x.x gateway were still arriving at the source host's physical uplink. Because the VM no longer resided on this host, these packets were naturally dropped.
Environment
VMware vSphere ESXi
VMware vCenter Server
Cause
A stale MAC address table (CAM table) within the physical switch fabric delays or drops the Reverse ARP (RARP) packet from the destination host, causing return traffic to route to the source host instead of the new destination host.
Resolution
This is not VMware/ Broadcom network stack issue. This is purely physical switch related issue. In order to resolve, following should be verified at physical switch fabric.
- Engage the physical network administration team to review the physical switch fabric. Ensure, auto update of mac-address table post each changes.
- Ensure all switch ports connected to the ESXi hosts are configured with Spanning Tree Protocol (STP) PortFast (or PortFast Trunk / Edge Port) to prevent listening/learning grace period delays.
- Verify if there is a synchronization delay across inter-switch links when updating shared MAC address tables in clustered switch environments (e.g., Cisco vPC, VSS, or stacked switches).
- Confirm that "Sticky MAC" or port security features are disabled on the ESXi-facing switch ports.
- Verify if Control Plane Policing (CoPP) or aggressive broadcast rate-limiting is configured on the switches and dropping the ESXi host's RARP broadcasts.
Additional Information
Feedback
