Failure scenario in vSAN stretched where ISL fails

[daphnissov]

After reading through the documentation, I'm not exactly clear on the failure behavior in the scenario where the ISL between the two data sites goes down but the witness still has access to both data sites. In this scenario, does the Primary/Preferred site win and all VMs are powered off at the Secondary and restarted on Primary/Preferred?

[daphnissov]

After some further discussions, it seems that upon failure of the ISL in which both data sites are partitioned from each other, the Preferred site takes precedence and VMs on the Secondary site will be powered off and restarted on the Preferred site. This behavior is roughly analogous to a failure of the Secondary site.