Diagnostics for VMware Cloud Foundation: vCenter Health: vSphere High Availability - Agent UnInitialization error

Products

• VCF Operations

Issue/Introduction

The error "vSphere HA agent is Uninitialized" indicates that the Fault Domain Manager (FDM) is present on the host but has failed to enter an operational state (Primary/Master or Secondary/Slave). Unlike an initialization error, this often happens on hosts that were previously healthy but lost their HA configuration or suffered a service crash.

Environment

VMware Cloud Foundation Operations 9.1

vSphere 7.x / 8.x / 9.x

Cause

FDM Service Crash: The FDM process on the ESXi host has stopped unexpectedly and the watchdog cannot restart it.

Management Agent Issues: Problems with hostd or vpxa agents on the ESXi host prevent vCenter from communicating with the HA agent.

Corrupted Configuration File: The local HA configuration file on the host's datastore or ramdisk is corrupted.

Network Isolation during Runtime: The host became isolated and failed to re-join the HA cluster after the isolation event ended.

Resolution

1. Restart Management Agents (Non-disruptive to VMs)

Often the HA agent is fine, but vCenter cannot see it because the management agents are stuck:

SSH to the host and run:

services.sh restart

Wait 2-3 minutes for the host to reconnect in vCenter.

 

2. Force Re-protection (Reconfigure)

This is the most effective way to "wake up" an uninitialized agent:

Right-click the ESXi host > vSphere HA > Reconfigure for vSphere HA.

Monitor the "Recent Tasks" – if it fails at 90%, it usually points to a network or VIB issue.

 

3. Clear Latent FDM State (Deep Clean)

If reconfiguration fails, the host may have a "stuck" state. You need to manually remove the agent:

Put the host in Maintenance Mode.

SSH to the host and remove the VIB:

esxcli software vib remove -n vmware-fdm

Reboot the host (optional but recommended to clear /tmp and memory).

Exit Maintenance Mode and Reconfigure for vSphere HA.

 

4. Check Datastore Heartbeating

Ensure the host has access to the designated heartbeat datastores. If a host loses access to storage, the HA agent may become uninitialized.

Additional Information

Logs for Troubleshooting:

ESXi Host: /var/run/log/fdm.log (look for UninitializedState or GenericFdmFault)

ESXi Host: /var/run/log/hostd.log (to check if management agents are responding)

VCF Tip: If this occurs during a VCF "Workload Domain" creation, ensure that the MTU settings (Jumbo Frames) are consistent across all hosts in the cluster.