VSAN HA critical fail/inaccessible VMs, Due to mutiple vSAN stretched clusters sharing the same inter-site link and witness appliances are hosted on those clusters.

Products

• VMware vCenter Server

Issue/Introduction

Mutiple vSAN clusters suffer HA events and VMs showing as inaccessible after inter-site link fails.  The vSAN clusters are hosting each others witness appliance and share the same inter-site link. 

Environment

VMware vSAN (all versions)

Cause

Per the Stretched cluster guide on page 7 in the Witness Host Location. 

A witness host appliance may run on another vSAN stretched cluster, but only if that other stretched cluster does not reside in the same sites for the cluster the witness host appliance is serving. This will help avoid an unintended situation in which one site failure initiates another site failure for another stretched cluster.

Resolution

Per the Stretched cluster guide on page 7 Witness Host Location section. 

The witness host appliance must reside at a third, tertiary location that is not the location of either one of the data sites for the stretched cluster it is associated with.