Hi,

Let’s say that backup of ovirt engine is done once every day, for example 6 AM using engine-backup tool.

Then during the day the changes are made that not only influence the database, but actually influence Libvirt/KVM/QEMU

(create/delete VMs,snapshots, disks, ).

Now let’s say at 18:00 an engineer restores from the backup.

How the ovirt will be ‘synchronized’ to Libvirt (/QEMU)?

Of course, it is OK to expect that system is being rolled back, but in that scenario there is nothing that I can imagine that will synchronize

Ovirt database state to Libvirt/KVM/QEMU state.

And once anything on (migrate/stop/start) – it will just fail in many cases because of inconsistency.

In 4.2 documentation

https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.2/html/self-hosted_engine_guide/reinstalling_hosts_she_backup_restore

there is sentence like:

“If migration is enabled at cluster level, virtual machines are automatically migrated to another host in the cluster; as a result, it is recommended that host reinstalls are performed at a time when the host’s usage is relatively low.“

Let’s say that there are 5 hosts (numbered from 1 to 5), 3 of them are hosted-engine (let’s say 1 to 3).

Then first of all, person should find a brand new host (host 6), then install HE on that. OK, I understand that HE can be then reinstalled on hosts 1 to 3.

But it when it comes VMs- if really migration of VMs can work, considering inconsistency above?

Eg. snapshot that is already in QEMU (done at 11:00) and the volume chain is updated is gone from database (because it is restored to 6 AM)

And in general, anything should be done towards non-HE hosts?

So far the restore seems to be very intrusive and quite error-prone in two parts: first part is actual handling of VMs.

And second part when ovirt-engine gets configured practically from scratch (through that dialog)

BR,

Konstantin