Connection interrupt during backup

[x1234]

Hi,

following problem occured today (for the first time, should be unrelated to PVE/PBS version):

When running a VM backup to PBS the connection between PVE and PBS got interrupted. However the Backup job still continued to run.
This caused the running VM that was being backupped to be stalling, I assume due to the "copy-before-write" filter in its block layer.

This of course causes troubles since the VM was completely blocked until I manually stopped the backup.

1) is there a setting that I overlooked that configures a timeout in such case? If not, any ideas how to react on this behaviour properly?
2) is there a solution other than cancelling the job? I assume the stalling can ultimately damage the VM which luckily seems not to have happened in this case


Thanks a lot for the great product(s)!

 

[Onslow]

Welcome, @x1234
Try "fleecing". There is a thread with a similar case:

Post in thread 'Failed Backup to PBS damages VM'

Tuesday at 10:23

Hi, please enable backup fleecing https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_vm_backup_fleecing Most likely your VM is I/O starved due to a slow backup target.

• Chris

• 

 

[x1234]

Hi,

I tried fleecing and it seems a good mitigation. When the connection drops, the fleecing runs full at some point and the backup cancels. The VM stays responsive at all times.

Follow up question:
Is using a sole zfs dataset for this, setting sparse and quota the correct approach for it?

zfs create rpool/fleecing
zfs set quota=10G rpool/fleecing
pvesm add zfspool fleecing-zfs --pool rpool/fleecing --content images --sparse 1 --nodes pve-carbon2

In my tests this was working fine but I'm unsure if I setup the dataset for it correctly