Architectural Overview
I have a development environment that is comprised of five (5) physical machines. Three (3) of these machines (dacloud-vh01, dacloud-vh02, dacloud-vh03) are VMM (2012R2) host servers set up as a failover cluster. One (1) of these machines (dacloud-ss01) is configured as a NAS (iSCSI) server which is used as the required shared file system for the cluster. The last machine (dacloud-cs01)--included here just for completeness--is a utility server that handles a variety of services including being one (1) of the four (4) domain controllers for the environment. The first four (4) systems are in an AD OU called "DACLOUD Cluster Hosts".
Methodology Background
Since it feel like it is probably not the best of ideas for the filesystem upon which the majority of the virtual machines as well as the shared volumes used for a few SQL instances exist to suddenly disappear as a result of dacloud-ss01 being rebooted following an automated update, I have created a GPO policy specifically for the DACLOUD Cluster Hosts OU (called "Default DACLOUD Cluster Hosts") that modifies the Windows Update settings for the rest of the domain. The non-default settings for this OU specify that the server should check for updates but not download or install them, as I would like to install these updates myself, manually.
What's Actually Happening
As demonstrated by the below screenshot taken from a GP Results report within Group Policy Management, the non-default WU settings are being applied by the DACLOUD Cluster Hosts policy. Moreover, in the second screenshot taken from the Registry Editor on dacloud-ss01, we can see that the WU settings are being pushed into the registry as expected.
That said, every Sunday morning @ 3AM, dacloud-ss01 checks for updates and, if found (let's face it, they're always found), installs the updates and, if required (let's face it, a reboot is always required), reboots the server. This, of course, causes pretty much the entire world to come crashing down in the environment as the filesystem that pretty much everything relies upon vanishes for about 3 minutes as the server reboots.
I'm entirely stumped by this problem, so I would be grateful if someone could point me in the right direction before I lose (another) (critical) virtual machine or corrupt (another) (critical) database. Thank you.