I just had to restore large container and it took a lot of time. Have you thought about live restore feature? It works great with VMs.
In theory it shoud be possible. My thoughts:
- pbs client have "mount" feature, so it is possible to mount the backup and have all data available immediately
- pbs has "protect" feature, so we can make sure the particular backup will not "disappear"
- there are at least a couple of "overlayfs" and "migratefs" projects out there. So we could mount pbs backup on bottom and a newly created volume on top. This way we would have all data available immediately and they could be slowly restored in the background from pbs.
- i think such CT could be fully usable. As long as the layers are properly mounted - we should be able to start, stop, reboot, backup (even to the same pbs!) or even snapshot or migrate it.
- in theory the CT could stay in such state forever. On the other hand when using migratefs we could have all files restored after some time. So we could simply restart the CT with updated mounts then.
This would allow to minimize downtime to nearly zero. What do you think?
This would be great feature and I see at least two applications:
1. restore after crash
2. migration to another pve cluster. (Just connect both to the same pbs. Backup here and restore there. Almost no downtime!)
In theory it shoud be possible. My thoughts:
- pbs client have "mount" feature, so it is possible to mount the backup and have all data available immediately
- pbs has "protect" feature, so we can make sure the particular backup will not "disappear"
- there are at least a couple of "overlayfs" and "migratefs" projects out there. So we could mount pbs backup on bottom and a newly created volume on top. This way we would have all data available immediately and they could be slowly restored in the background from pbs.
- i think such CT could be fully usable. As long as the layers are properly mounted - we should be able to start, stop, reboot, backup (even to the same pbs!) or even snapshot or migrate it.
- in theory the CT could stay in such state forever. On the other hand when using migratefs we could have all files restored after some time. So we could simply restart the CT with updated mounts then.
This would allow to minimize downtime to nearly zero. What do you think?
This would be great feature and I see at least two applications:
1. restore after crash
2. migration to another pve cluster. (Just connect both to the same pbs. Backup here and restore there. Almost no downtime!)
Last edited:
