I am currently trialing Urbackup to backup some Hyper-V servers, and I am impressed with it.
I have also purchased the block tracking addon.
Just wondering, what would be the best strategy for backing up the virtual machines? I have currently told it to exclude all VHD and VHDX files from the backup, then told it to backup the VM’s under “Configure Components to backup” in the client app.
This works, but it copy’s the entire VHD/VHDX each time.
Is this recommended, or should I install the client in each VM, and let it do both an image and file backup of each?
I’m working on a special client for Hyper-V 2016, but that will be a bit yet. I’ll also write up a documentation page detailing and comparing backup methods when that is finished.
Backup via Windows Backup API (VSS) should work as well. Also in combination with CBT. The CBT works on a per-volume basis and does not work with something like Cluster Shared Volumes (CSV).
W.r.t. the vhd/vhdx files the following is important with the VSS backup method:
Read only the file differences on the Hyper-V server. CBT takes care of that.
Transfer only the differences during backup. See the advanced setting of the client (set incremental transfer mode to block hash differences)
Store only the differences on the server. Currently only possible with btrfs or ZFS as file systems (see the btrfs and ZFS manual sections for setup). Next major server version will support XFS (formatted with reflink support) and ReFS v2 as additional file systems that can store file differences, but those do not have file compression which is also nice to have
Then I would say that depending at what runs in the VM, running the client in the VM might be better. That way one can better exclude unnecessary data, and separate critical data that needs to be backed up often from data that where a backup is only a convenience for faster restores (like the default setup where an complete image backup is run every week and file backups of user data every 6h).
I use XenServer as a Hypervisor instead of Hyper-V, but i’ve given up on trying to backup the hypervisor itself. I’m only doing backups of the VM’s themselfs from clients installed in the server. Gives me the option to be a bit more granular with what I want to backup, or more to the point, what not to backup per VM.
Plus I like to keep the hypervisors themselfs as clean as possible, i’ve had issues with performance and stability on a hypervisor before, which resulted in a tug of war between us and Citrix about unsupported additions to the platform that were probably the cause of the issues… Wasn’t the issue in the end, but did cause us to do nothing to the hypervisors but install and update them.
I can see the appeal of creating a backup from Hyper-V though, especially if you rent out VPS and customers won’t want to install local backup clients…
