Sanity Check: Deploying a vSAN Witness Appliance

I recently stood up a 4-node vSAN Cluster (single site/non-stretched), and I was a little disappointed with how a policy of FTT=1 would require 3 functional hosts and that I couldn’t split the 4 nodes into 2 availability zones (hosts are distributed 2 per rack). Is it possible to deploy a vSAN Witness Appliance (to be hosted on another VMware cluster) in order to improve the number of tolerable host failures?

Based on what I’ve read about the witness appliance on a 2 node cluster, I fail to see why it wouldn’t. Especially since the “Large” Witness Appliance is capable of supporting up to 500 VMs.

I’m struggling to find documentation that either supports or squashes the idea. So any insight would be appreciated.

View Reddit by PacketDropperView Source


To see the full content, share this page by clicking one of the buttons below

Related Articles


  1. You could do a 2+2 stretched cluster.

    Why can’t vSAN operate without only 2 FD and no witness and so automated failover? This is a computer science problem. You can’t avoid split brains with 2 fault domains and no quorum…

  2. You’re essentially looking for a stretched cluster setup.

    That way you can create a fault domain ‘per site’ with 2 hosts and have a witness in between.

    2 + 2 as non stretched clusters cannot work because you’re not having a quorum in that case.
    So if network drops between the two fault domain, nobody’s gonna know who’s in charge.

Leave a Reply