Reboot = FullBackup?

[Janko]

Hi Folks,

everytime i reboot a VM PBS makes a a full Backup from the VM during the next Backup Job.
With a Cluster of over 200VM and with over 80TB Storage.... its really annoying and an performance Issue.

Is it possible to prevent it?

 

[tom]

Before you reboot a host, you should live migrate your VMs to a remaining host.

Live migrate will preserve the dirty-bitmap status.

 

[czechsys]

Before you reboot a host, you should live migrate your VMs to a remaining host.

Live migrate will preserve the dirty-bitmap status.

Live migrating require same CPU or default kvm64 which is slower due missing cpu flags...
And if pve reboot unplanned...

And btw, OP is meaning VM reboot, not pve. Just think about kernel update & reboot...

 

[fabian]

it will soon also be possible to save the bitmap when doing a 'hibernate' style VM suspension - the full machine state including guest memory and dirty bitmaps is persisted to disks in that case. that way you can also preserve it across PVE node reboots if desired.

 

[tom]

And btw, OP is meaning VM reboot, not pve. Just think about kernel update & reboot...

If you reboot a VM (e.g. after applying updates inside), the dirty-bitmap is still there as the KVM process is still running.

 

[czechsys]

If you reboot a VM (e.g. after applying updates inside), the dirty-bitmap is still there as the KVM process is still running.

And if you reboot VM via pve because you want change some non-hot-changeable parameter...

 

[tom]

And if you reboot VM via pve because you want change some non-hot-changeable parameter...

In this case, the backup process has to read all data again. But you do not do this for 200 VMs at the same time, just to quote the original thread starter.

 

[czechsys]

Really?

https://forum.proxmox.com/threads/e...d-the-vm-has-been-restarted.73175/post-326915

 

[tom]

Really?

https://forum.proxmox.com/threads/e...d-the-vm-has-been-restarted.73175/post-326915

I do not get your question, what is unclear?

 

[czechsys]

I am showing that there is a lot of use-cases when VM(s) are restarted from pve point of view. Doing full backup(s) in Always Incremental scheme just because there was VM(s) restart is bad way.

 

[fabian]

I am showing that there is a lot of use-cases when VM(s) are restarted from pve point of view. Doing full backup(s) in Always Incremental scheme just because there was VM(s) restart is bad way.

it's not doing "full backups" in any sense of that word. it just can't do a fast incremental backup, and has to do a regular incremental one instead. think selectively passing changed files only to rsync (because you somehow know only A, B and C are different from last time), vs. letting rsync determine which files have changed.