in Paused-Critical state, while VM4 continues to run. This is because VM3 is stored on
a CSV and therefore loses its storage, while VM4 uses SMB storage, which enables it
to continue running as its storage is still accessible.
Figure 7.19 Cluster network properties
Storage resiliency is also utilized in full-disk scenarios. In Windows Server 2012 R2,
when a volume has less than 200MB of free space, Hyper-V will pause any VMs with
dynamic/ differencing disks, while VMs using fixed-size VHDs are not affected. In
Windows Server 2016 that uses storage resiliency, a VM is paused only when
performing an I/O that will fail due to the VHD being unable to grow. Once a VM is
paused in Windows Server 2016, you will need to resume it manually after you have
made additional disk space available.
Shared VHDX behavior is slightly different, because multiple VMs are using the same
VHD spread over multiple nodes in the cluster. With Shared VHDX files, if a VM loses
connectivity to the Shared VHDX, instead of the VM being paused, the Shared VHDX
is removed from the VM. This allows the VM to notice that the disk is missing and
take appropriate action through the cluster capabilities; for example, moving the
resource using the disk to another node in the cluster that does have access. Any VM
that loses connectivity will check every 10 minutes for the storage to be re-attached
(when the Hyper-V host regains access to the underlying storage).
Storage resiliency works for pretty much every deployment combination that utilizes
CSV (either directly or through a Scale-out File Server), whether it’s a generation 1 or
generation 2 VM and whether the disk is VHD or VHDX. Only VMs using local storage