vCenter performance degradation when legacy dVS is present

Products

• VMware vSphere ESXi

Issue/Introduction

Customers may experience severe vCenter Server performance degradation, slow vSphere Client response times, or timeouts when executing management tasks. This condition frequently occurs in environments that have undergone an NSX-V to NSX-T migration where legacy Distributed Virtual Switches (dVS) with NSX-V tie-ins (such as vxw-xxxx port groups) remain attached to ESXi hosts alongside the new NSX-T dVS configuration.

Environment

• vCenter Server 8.x

• VMware ESXi 8.x

• VMware NSX-V (legacy)

• VMware NSX-T  4.x

Cause

During bulk VM provisioning operations (such as deploying Horizon VDI desktops), vCenter attempts to resolve and evaluate valid network placement paths across all attached switches.

When legacy NSX-V dVS objects remain attached to hosts without active vNICs or VMkernel adapters, vCenter becomes confused by the dual-dVS configuration. This triggers severe background reconciliation overhead:

• VPXD Service Saturation: vCenter continually attempts to poll, sync, and reconcile stale network topologies across hosts. This spikes VPXD thread usage and delays inventory refreshes.

• Database & Lock Contention: Ghost objects trigger persistent database sync attempts, cluttering tables (VPX_DVPORT, VPX_DVPORTGROUP) with conflicting state data.

• API & UI Responsiveness: UI tasks freeze while vCenter attempts to map path bindings or resolve missing port group IDs.

 

Resolution

• Navigate to Inventory > Networking.

• Locate the legacy NSX-V dVS (verify it has no active uplinks, physical NICs, or VMkernel interface mappings).

• Right-click the legacy switch and select Add and Remove Hosts...

• Select Remove hosts, pick all affected ESXi hosts in the cluster, and complete the wizard.

• Once all hosts are detached, right-click the stale dVS and select Delete.