Hyper-V CSV backup: What needs to be considered for VM backups?

The following points should be observed when backing up Hyper-V VMs to CSVs.

1. The latest version of BackupChain should be used
2. All VM files must be saved on the same CSV
3. The backup should only be done via the Hyper-V tab. Complete server image backup must not include CSV volumes; it should only include the system OS and optionally data disks.
4. If it is likely that VMs will be moved to other nodes, the auto-select feature should be used instead of selecting the VMs from the list A speed limit is activated when the cluster option is set when the task is created. This can be increased or switched off if you are sure that CSV management traffic is 100% isolated and cannot be affected by backup traffic and other data transfers. Otherwise it can happen that the heartbeats do not arrive in time and Hyper-V completely switches off the node

As with all Hyper-V backups, depending on the host and guest OS, it may happen that Hyper-V “slides in” a tiny checkpoint shortly before the backup, which is deleted immediately after the init phase. This checkpoint file appears in the backup folders as * .AVHDX. In addition, you will also find other AVHD / X in the backup folders if the VM has other checkpoints.

We recommend that checkpoints not be used and if you use them, they should only be used for a short time. When using checkpoints, there are also some disadvantages and risks to consider. Granular recovery in BackupChain only works on VHD / X basis. Checkpoints cannot be inspected with that feature, i.e. a complete recovery is more likely necessary if the desired files cannot be found in the main VHD file. When you create a checkpoint in Hyper-V, the VHD is frozen (and previous checkpoints are, too) and a new AVHD/X is created. All changes within the VM will in future be written to the AVHD on a sector basis. When deleting the checkpoint, the contents of the AVHDs must be merged with the parent VHD, which can take a while.

The disadvantages are higher complexity and slower hosts and network access. Risks are possible data loss, there were already several bugs in Hyper-V that led to complete loss. A few risky phases have remained, for example the merge process when a checkpoint is deleted or when VHD references are set up when checkpoints are created. If, for example, the system is backed up or there is a power failure, the memory of the CSV is also separate and additionally affected, there is a risk of data corruption. Microsoft has of course improved the VHDX format to reduce risks, but virtual hard drives are certainly not 100% protected against corruption from all causes.