Hi Sean. Sorry for the late reply. What we want to do is backing up snapshots in case of a complete compute lost of as a plan for disaster recovery. So after recreating the environment we can restore snapshots and start the VMs again. Cheers On Wed, Mar 4, 2020 at 10:14 PM Sean McGinnis <sean.mcginnis@gmx.com> wrote:
On 3/4/20 9:58 AM, Gorka Eguileor wrote:
On 03/03, Alfredo De Luca wrote:
Hi all. We have our env with Openstack (Train) and cinder with CEPH (nautilus) backend. We are creating automatic volumes snapshots and now we'd like to export them as a backup/restore plan. After exporting the snapshots we will use Acronis as backup tool.
I couldn't find the right steps/commands to exports the snapshots. Any info? Cheers
-- *Alfredo* Hi Alfredo,
What kind of backup/restore plan do you have planned?
Because snapshots are not meant to be used in a Disaster Recovery backup/restore plan, so the only thing available are the manage/unmanage commands.
These commands are meant to add an existing volume/snapshots into Cinder together, not to unmanage/manage them independently.
For example, you wouldn't be able to manage a snapshot if the volume is not already managed. Also unmanaging the snapshot would block the deletion of the RBD volume itself.
Cheers, Gorka.
If the intent is to use the snapshots as a source to backup the volume data, leaving the actual volume attached and IO running but still getting a "static" view of the code, then you would need to create a volume from the chosen snapshot, mount that volume somewhere that is accessible to your backup software, perform the copy of the data, then delete the volume when complete.
I haven't used Acronis myself, but the issue for some backup software could be that the volume it is backing up from is going to be different every time. Though you could make sure it is mounted at the same place so the backup software at least *thinks* it's backing up the same thing.
Then restoring the data will likely require some manual intervention, but that's pretty much always the case when something goes wrong. I would just recommend you test the full disaster recovery scenario to make sure you have that figured out and working right before you actually need it.
Sean
-- *Alfredo*