vSAN Datastore capacity incorrect due to cluster partition
search cancel

vSAN Datastore capacity incorrect due to cluster partition

book

Article ID: 441501

calendar_today

Updated On:

Products

VMware vSAN

Issue/Introduction

This article addresses a scenario where the vSAN datastore reports a significantly lower total capacity than the combined physical disks across all ESXi hosts. This usually occurs when a network partition prevents hosts from contributing their storage to the global datastore.

Symptoms

  • The vSAN datastore reports a total capacity equal to the disks on a single host (e.g., 31 TB) instead of the combined total of all hosts (e.g., 90 TB).
  • vSAN Skyline Health indicates a cluster partition where each ESXi host is isolated in its own partition.
  • The unicast agent list on the ESXi hosts is empty, showing no peer entries.
  • Virtual machine objects may show as inaccessible or reduced-availability-with-no-rebuild.
  • Attempts to ping peer vSAN VMkernel interfaces with large packets (vmkping) may fail.

Environment

ESXi version 8.0 u3

Cause

A cluster partition is occurring because the ESXi hosts cannot establish vSAN network communication, often due to an empty or incorrect unicast agent list.

When a partition exists, vCenter Server displays capacity statistics from only one partition at a time, resulting in a size mismatch.

Resolution

  1. Verify Cluster State: Run the following command on each host to check the partition status:
    localcli vsan cluster get
    Review the Sub-Cluster Member Count. If the count is 1, the host is isolated.

    Example of localcli vsan cluster get output during a partition:

    Cluster Information:   
    Enabled: true
    Local Node State: MASTER
    Sub-Cluster Member Count: 1
    Sub-Cluster Member HostNames: [HOSTNAME]


  2. Validate Network Connectivity: Test the vSAN VMkernel port connectivity using vmkping with the MTU size used for vSAN (e.g., 8972 for MTU 9000):
    vmkping -I vmkX [IP_ADDRESS] -d -s 8972


  3. Rebuild Unicast Agent List: Restore the unicast agent list on all affected hosts to enable peer discovery. Refer to Configuring vSAN Unicast networking from the command line for detailed steps.


  4. Confirm Resolution: Once the lists are rebuilt, verify the Sub-Cluster Member count has increased:
    esxcli vsan cluster get

     

The vSAN datastore capacity should now reflect the combined total of all contributing hosts.