On
Fri, Mar 4, 2022 at 8:28 AM Vinícius Ferrão via Users
<users@ovirt.org>
wrote:
Hi again, I don’t know if it will be possible to import the storage domain due to conflicts with the UUID of the LVM devices. I’ve tried to issue a vgimportclone to chance the UUIDs and import the volume but it still does not shows up on oVirt.
LVM
can change VG/PV UUIDs and names, but storage domain metadata kept in the
VG
tags and volume metadata area contain the old VG and PV names and UUIDs,
so
it is unlikely to work.
The
system is designed so if the original PVs are bad, you can
disconnect
them and
connect
a backup of the PVs, and import the storage domain again to the system.
Can
you explain in more details what you are trying to do?
Nir
What I was trying to accomplish is get some VM snapshots from days ago.
On my storage system it generates a snapshot of the entire disk pool in a daly basis. So it was natural, at least for me, to mount a snapshot at a given time to export some VMs from those snapshots. The environment had an attack and needed to be recovered
from it, and rolling back was the approach.
But that what the issue happened, when trying to import the snapshot from the iSCSI share it conflicted with the running iSCSI share, since it’s a storage snapshot, it have the exactly same metadata.
I think the issue here is a missing feature on oVirt to remap the metadata and permit the mount in this situation, right?
What I ended up doing:
* Removed one of the servers from oVirt Cluster / Datacenter
* Reinstalled from the ground
* Fired up a new hosted engine on a new iSCSI HE dedicated share
* Reconfigured everything: network, VLAN, iSCSI, etc.
* == IMPORTED the snapshot on this new engine
* Created an export domain on another NFS share
* Exported the 6 VM’s that I need to export from the storage-level snapshot to the export domain
* Detached the export domain from the temporary engine
* Added the export domain on the production engine
* Deleted the compromised VMs
* Imported the “snapshoted" VM’s from the export domain
* == INFRA is back
* Destroyed the new engine, the snapshot iSCSI share and the export domain
* Reinstalled the host
* Added back to the original Datacenter / Cluster.
As you can see was a tiresome work just to get the VM’s back from an storage level snapshot of the iSCSI share, but is what I’ve ended up doing.
Lesson learned: it’s too difficult to recover storage-level snapshots, it’s hard on NFS and on iSCSI is extremely worse since you can’t mount iSCSI whatever you want to.
My opinion: Should be a feature on oVirt to at least allow to mount this kind of conflicting volume as readonly for recovery reasons only.
Thanks.