DRS fails to evenly load balance virtual machines following a vSphere HA reconfiguration

Products

• VMware vCenter Server
• VMware vSphere ESXi

Issue/Introduction

• After reconfiguring vSphere HA on a cluster, vSphere DRS is not balancing the virtual machines across all available hosts, resulting in an uneven resource distribution.
• A subset of hosts is running at 100% CPU and Memory utilization.
• All virtual machines are hosted on a few hosts, leaving others idle or significantly underutilized.

Cause

The load imbalance occurred because of network connectivity issues between the vCenter Server and a subset of ESXi hosts. In most cases, this results from stale or incorrect DNS records that prevent proper name resolution. When vSphere HA was reconfigured, the vCenter Server attempted to communicate with all hosts but was unable to reach the affected systems, leading to the imbalance.

Resolution

To resolve this issue, 

1. Identify and remove the duplicate IPs for the affected ESXi hosts
2. The DNS cache must then be cleared to restore stable network communication with the vCenter.
   • Restart the dnsmasq service on the vCenter to clear the local DNS cache: Clear DNS Cache on vCenter Server

Once connectivity is successfully restored between all ESXi hosts and vCenter, DRS will automatically detect the load imbalance and begin migrating VMs to rebalance the CPU and Memory utilization across all available hosts in the cluster.