Now and then a thread pops up where someone has problems with a backup job were backups of some of the VMs fails. I have investigated this problem and I might have found the cause for this problem.
The problems boils down to two specific issues:
Why is the above issues a problem?
The problem is that the backup job only serializes backups on one Proxmox node which means that if your backup job contains VMs from more than one node the backup job will start a backup from the backup list on each Proxmox node in parallel. Expressed mathematically f(x) = n * b(p); n = Proxmox nodes; b = backup from node p.
How can I overcome this problem?
To prevent running parallel backups create a backup job for each Proxmox node and ensure the backup job only contains VMs which runs on the same proxmox node.
Is this bullet proof?
Yes and no.
Yes: If you always backup every VMs then it is bullet proof if you use selection mode: all
No: If you only want to backup running VMs then you will have a problem since over time the list of running VMs on a given Proxmox node will eventually change due to migration, new VMs can be added, VMs can be removed, VMs can be stopped so this lists needs to be overlooked on a daily basis.
How can this limitation be solved?
Add a new selection mode called running which well select all VMs on the configured node with status running.
The problems boils down to two specific issues:
- Limited IO in either the backup storage or the network
- The configured backup job contains VMs from more than one Proxmox node
Why is the above issues a problem?
The problem is that the backup job only serializes backups on one Proxmox node which means that if your backup job contains VMs from more than one node the backup job will start a backup from the backup list on each Proxmox node in parallel. Expressed mathematically f(x) = n * b(p); n = Proxmox nodes; b = backup from node p.
How can I overcome this problem?
To prevent running parallel backups create a backup job for each Proxmox node and ensure the backup job only contains VMs which runs on the same proxmox node.
Is this bullet proof?
Yes and no.
Yes: If you always backup every VMs then it is bullet proof if you use selection mode: all
No: If you only want to backup running VMs then you will have a problem since over time the list of running VMs on a given Proxmox node will eventually change due to migration, new VMs can be added, VMs can be removed, VMs can be stopped so this lists needs to be overlooked on a daily basis.
How can this limitation be solved?
Add a new selection mode called running which well select all VMs on the configured node with status running.
Last edited:
