Repository navigation
Hyper-V Questions #10
|
The addition of support for resilient change tracking across Hyper-V clusters piqued my interest. A few questions, if you have time.
Thank you. |
Replies: 2 comments 5 replies
|
Thanks for the questions! Application-aware only: today BackupZit first takes an application-consistent recovery checkpoint (VSS inside the guest). If that fails, it takes a crash-consistent one and the run ends with a warning that says so, so it's visible but the job doesn't fail. There's no "fail instead of falling back" switch yet. The next release (0.48) adds a per-job option "Require application-consistent checkpoints": with it on, the run fails if the guest checkpoint isn't application-consistent, with no fallback. Inventory: the agent on every node lists the VMs on that node (Get-VM) and reports them to the console about once a minute. The console groups the nodes by cluster name, and each VM is backed up by the agent of the node that owns it at that moment; the other nodes skip it. Collecting the inventory is read-only, so live-migrating a guest meanwhile is fine: at worst it shows on the old node until the next report. Live migration during the backup of that VM is something we'll test on our lab cluster before 0.48, and we'll fix whatever we find (e.g. cleaning up the checkpoint through the cluster on the current owner node). Until then I'd keep migrations out of the backup window. Shielded VMs / vTPM: shielded (HGS-guarded) VMs aren't supported yet. For vTPM: the backup stores the VM's configuration including "TPM enabled". Restoring over the existing VM keeps that VM, and with it its vTPM and key protector. Restoring as a new VM (or after the original was deleted) currently creates a new local key protector and an empty vTPM, so a guest with BitLocker will ask for its recovery key. We're working on keeping the vTPM state when restoring to the same host or cluster, and I'll post here when it's ready. Until then, keep your BitLocker recovery keys. Thanks for raising these: questions like this help us make the Hyper-V support better. |
|
Update: all three are ready for testing in the release candidate BackupZit 0.48.1 (release notes). We tested them on our lab cluster (two Windows Server 2022 nodes, CSV over iSCSI), but our test VMs are Linux and empty Windows guests, so a test with your AD/SQL VMs would be very valuable. If you have time, here is how: Install the candidate (it is not the default release yet)
1. Application-consistent only
2. Live migration during a backup
3. vTPM / BitLocker
Please reply here with what worked and what didn't (the run messages help). Thanks a lot! |
Thanks for the questions!
Application-aware only: today BackupZit first takes an application-consistent recovery checkpoint (VSS inside the guest). If that fails, it takes a crash-consistent one and the run ends with a warning that says so, so it's visible but the job doesn't fail. There's no "fail instead of falling back" switch yet. The next release (0.48) adds a per-job option "Require application-consistent checkpoints": with it on, the run fails if the guest checkpoint isn't application-consistent, with no fallback.
Inventory: the agent on every node lists the VMs on that node (Get-VM) and reports them to the console about once a minute. The console groups the nodes by cluster name, …