2-node stretched cluster shows vSAN Cluster Partition error in Skyline Health after upgrading to vSphere 8.

Products

• VMware vSAN

Issue/Introduction

Symptoms: 

• vSAN Skyline Health show an error: "vSAN Health Service - Network Health - vSAN Cluster Partition"
• When performing  Cluster -> Monitor -> vSAN -> Virtual Objects you observe that many or all of the objects are in "Reduced Availability Mode with no rebuild" status
• The witness is shared with other clusters

Environment

• VMware vSAN 8.x

Cause

• The witness size profile is too small for the environment, so the "Arena" memory heap on the witness has become exhausted.

 

Verification:

• You can see entries similar to the following in the vmkernel.log on the ESXi host(s) around the time the stretched cluster is being configured:

2025-07-15T23:47:36.575Z cpu1:6304289)CMMDS: CMMDSLogStateTransition:1917: XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX: Transitioning(XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX) from Rejoin to Standalone: (Reason: Arena memory exhausted)

• From an ssh session you can run the following command verify that the epd service is running:

/etc/init.d/epd status

 

If the service is not running, please see KB: vSAN cluster Partition with error " Arena memory exhausted"

Resolution

Workaround:

• Deploy a new witness appliance using an increased size profile and reconfigure affected clusters to use the new appliance. See KB: vSAN Stretch Clusters - How to Replace the Witness Appliance/Host

Additional Information

• NOTE: This may initially seem similar to KB: vSAN cluster Partition with error " Arena memory exhausted", but the epd service is still running in this case.  Furthermore, the DISCARDED_COMPONENT count may not be high in the case, as well. Please contact Broadcom support for assistance with confirmation of service status and determining the DISCARDED_COMPONENT counts, if needed.
  
  
• For more information on deploying the vSAN witness appliance and size profiles, see TechDocs: Deploying a vSAN Witness Appliance