When you’re dealing with Windows Server 2012 R2 CSV backups you need to be aware of the following:
- use Hyper-V Backup types only, not Universal
- overlapping backups (simultaneous backups between two or more nodes) are now permitted. Server 2008 R2 couldn’t do that
- make sure all VM files are on the one and same CSV: checkpoint files, VHDX, etc. Don’t spread files for one VM across multiple CSVs and don’t use local disk folders
- Do not use the Folder based backup method, only the Hyper-V tab in BackupChain, which automatically updates itself with the most current VM configuration each time the backup runs (See User Guide Tutorials section)
- each node can only backup VMs running on the same node. That’s why you need a BackupChain instance on each node, too. It’s a CSV architectural requirement created by Microsoft, not us.
- when restoring a VM, restore it first to a local temp folder, then use the Clustering screens in Windows to move it into a CSV
- Speed limit (Speed tab): when you create a Hyper-V backup in BC and check the option ‘cluster’ in the New Task Wizard, BC adds a backup speed limit. Some CSVs out there aren’t set up according to Microsoft’s requirements and the entire cluster may crash when a backup starts up. When this occurs, it’s almost always due to delays in the heartbeat signal. When it takes too long, the cluster thinks the node is gone and tears it down.
Speed limits aren’t really necessary, but some hardware networks out there aren’t set up correctly and thus don’t isolate backup traffic from data and cluster traffic. When the backup traffic spikes, the cluster traffic gets delayed, which should never happen.
The speed limit may be increased or lifted in steps when you are certain everything has been set up correctly, ideally during a time when you could risk a tear down without losing productivity.
If your network and CSV were set up correctly there is nothing to worry about and there is no need to keep the speed limit unless you want to reduce the traffic on your network.