BackupZit 0.48 brings application-consistent Hyper-V backups, vTPM support, restores over the REST API with an OpenAPI specification, and a series of fixes found and contributed by Ermo75 — thank you! It contains everything from 0.47.4.
Fixes contributed by Ermo75
- vTPM state kept by maintenance: prune, backup copy jobs, immutability extension and the repository check also cover the vTPM state of Hyper-V backups (#12). The check reports a missing state.
- Overlapping backups of the same Hyper-V VM wait for each other instead of removing each other's checkpoint (#18). A checkpoint left behind by a crashed backup is removed automatically after 48 hours.
- Restoring over an existing Hyper-V VM first builds all disks and only then stops the VM. If reading the backup fails, the VM keeps running unchanged (#20).
backupzit-agent hyperv restoreholds a repository lock, like restores from the console (#14).- The agent log says why a Hyper-V disk was read in full instead of only its changes, and how much was read (#22).
Also updated: minio-go, sftp and govmomi.
Hyper-V
From the questions in discussion [#10]:
- Require application-consistent checkpoints (new job option, also for jobs of a failover cluster): a running VM is backed up only from a checkpoint the guest prepared with VSS. If that is not possible, the VM is not backed up and the run fails (with a notification) instead of taking a crash-consistent checkpoint. The other VMs of the job are still backed up and can be restored from that run. Meant for Active Directory, SQL Server, Exchange and other databases. Needs agents 0.48.0 or newer; older agents get a failed run asking to update them.
- Live migration during a backup: a VM that moves to another node while it is being backed up is still backed up completely. Until now the backup's checkpoint stayed on the VM (silently) until the next backup; now the node the VM moved to removes it within a few minutes, and the run says that the VM moved.
- vTPM and BitLocker: backups of Generation 2 VMs with a virtual TPM also keep the vTPM (guest state file and key protector). Restored as a new VM on the same host, or on a host with the original host's guardian certificate, the VM gets its vTPM back and BitLocker does not ask for the recovery key. On other hosts the VM gets a new, empty vTPM and the restore explains how to avoid that. Shielded VMs (Host Guardian Service) are not supported.
- A restore on a cluster node adds the VM as a cluster role only when the VM was clustered or its files are on a Cluster Shared Volume; before, a restored local VM also became a role that could not fail over.
REST API and webhooks
- OpenAPI specification: the console serves it at
/api/v1/openapi.json(no token needed); the file is also in the repository: internal/server/openapi.json (github.com). A test keeps it in step with the code. - Restore over the API:
POST /api/v1/runs/{id}/restorerestores files of a backup into a folder on an agent, or to the original location (role: restore operator; in the audit log). GET /api/v1/runs/{id}also returns the output of the commands before and after a backup.-
Webhooks (
run.finished) now also carryagent_id,duration_s,files,bytes,bytes_read,bytes_storedandsnapshot, so a dashboard needs no extra API call. -
Browsing and downloading files, disk images or VM disks of a backup made shortly after an earlier browse no longer fails; a download that cannot be read now shows a message instead of an empty zip.
Guides updated in all 14 languages.
Releases are signed with the BackupZit release key; consoles verify them before installing.