We've been struggling to create the consistent scheme for backup of KVM virtual machines under Proxmox control and Ceph as storage, but still quite often during the vzdump with a snapshot the virtual machine is going down
120: dic 05 04:51:47 INFO: status: 17% (21731147776/127775277056), sparse 0% (1095598080), duration 1225, 17/16 MB/s
120: dic 05 04:52:47 ERROR: VM 120 not running
120: dic 05 04:52:47 INFO: aborting backup job
120: dic 05 04:52:47 ERROR: VM 120 not running
120: dic 05 04:52:48 ERROR: Backup of VM 120 failed - VM 120 not running
Vzdump config
vzdump 120 --quiet 1 --mode snapshot --mailto <email> --node node1 --compress lzo --maxfiles 2 --storage backup01
Some of our customer proposed us the following strategy of the backup from the inside of the virtual machine, saying "it is that simple!"
mount /nfs/backup
dd if=/dev/sda of=/nfs/backup/$datetime-$virtualmachine
What would be your opinion on this proposal? How to explain in a simple and profound terms that this is a really bad idea? And how to remain calm and polite in this situation:?
Thank you!
No, this is not going to work because the disk
/dev/sda
will be changing during thedd
run and you end up with an inconsistent image.Simpler explanation: Tell him that if you create a file towards the end of the backup, it will not be be backed up as the information where a file is located is at the beginning of the image and will not include the new file.
Better yet: Just demonstrate it. It's easy to do and the result is usually spectacular...