Getting Unicast messages in the vmkernel log - Failed to send to unicast host 'x.x.x.x;12321' on iface 'y.y.y.y': Host is down.

Products

• VMware vSAN

Issue/Introduction

We have unicast errors in the vmkernel log between the vSAN and Witness network. But the vSAN environment is working fine.

Below are examples of what is seen in the environment

running the esxcli vsan cluster get command shows on each cluster node shows it to be healthy:

Cluster Information:
   Enabled: true
   Current Local Time: 2025-10-28T21:08:28Z
   Local Node UUID: local_node_id
   Local Node Type: NORMAL
   Local Node State: MASTER
   Local Node Health State: HEALTHY

 

And vSAN witness tagging appears to be correct:

 Interface  NetstackInstance   Tags
    ---------  ----------------   ----
    vmk0       defaultTcpipStack  VSANWitness, Management
    vmk1       defaultTcpipStack  VSAN
    vmk2       defaultTcpipStack  blank in esxcfg-info_-a.txt
    vmk3       vmotion            VMotion

Environment

vCenter 8

vSAN 8

ESXi 8

Cause

It appears that there may be an issue with the tagging that causes the witness traffic to go to a different route

Resolution

Place each host in the cluster in maintenance mode, one at a time, and then remove the vSAN witness tag from the vmk port and save and then re-tag the vmk port and save