ESXi host is marked as "not responding" in the vCenter Server upon physical switch ports corresponding to the host's iSCSI uplinks are shut down.

Products

• VMware vCenter Server

Issue/Introduction

• The ESXi host is getting into not responding state on vCenter upon shutting down the physical switch interfaces connected to ESXi iSCSI vmkernel VMNIC.
  
  

• During the "not responding" state, it is confirmed that the host is still online and fully functional. The host's UI is successfully accessed, proving the management network (vmk0) is not affected.

Environment

VMware vSphere ESXi

Cause

• In the affected cluster, iSCSI vmkernels are in following IP range (142.40.x.x):
  
  
  
• Upon reviewing the vCenter network configuration. The IP address of eth0 interface is also same network subnet: (142.40.x.x).
  
  
• The above misconfiguration of setting the ESXi iSCSI vmkernels and vCenter IP in same network range is causing the exchange of UDP heartbeat packets to go through the iSCSI VMKernels (VMK2), instead of using the expected VMKernel (VMK0). 
  
  
• Hence, whenever the uplink switch interface connecting to iSCSI VMKernel VMK2 is shut, the heartbeat exchange between host and vCenter  stops. Please find the details in the following vCenter vpxd.log:
  
  
  2025-11-13T10:13:28.499+08:00 info vpxd[]  sub=HostCnx opID=CheckforMissingHeartbeats-520a712c] [VpxdHostCnx] No heartbeats received from host; h: host-2815, time since last heartbeat: 148421ms
2025-11-13T10:30:40.510+08:00 info vpxd[] [ sub=InvtHostCnx opID=HeartbeatStartHandler-51d089da] Missed heartbeats for host; [vim.HostSystem], missed: 117,

Resolution

The following steps should be followed to resolve this issue:

• By moving the vCenter IP and iSCSI vmkernel IPs to a separate different subnet.

• This change forces all vCenter communication, including the critical UDP heartbeats, to correctly use the designated ESXi management VMkernel (vmk0), as intended.

• Once heartbeats are flowing correctly over the stable management network, shutting down the iSCSI switch ports will no longer interrupt them. This prevents vCenter from falsely marking the host as "not responding."

• This solution aligns with VMware's fundamental design principle of network isolation. Management, vMotion, and storage traffic (like iSCSI) should always be on separate, non-overlapping IP subnets to ensure predictable routing and prevent interference.

Additional Information

Note: Kindly note the above issue is applicable to any other VMKernel configured on the ESXi host in the same IP subnet as the vCenter appliance.