[PVE-User] Backup server: Backup not finished because of network outage, can failing backup potentially cause read-only root FS?

[Benjamin Hofer <benjamin@gridscale.io>]

Dear Proxmox community,

we had a 13 minutes firewall outage. During that time, there was an hourly
VM backup running.
The backup didn't finish as the Proxmox VE node couldn't reach the backup
server.
I see the "GET" API request (return code 101) that starts the backup in the
PBS API logs. The HTTP "PUT" to transfer the backup is missing.
Unfortunately the backup's task log is missing. It obviously got rotated.
During that time, the root filesystem of the VM changed to read-only.

To find out if things are related or not, it'll be helpful for me to answer
the following questions:

1) What happens when a Proxmox backup is started and PBS is not reachable
anymore to transfer the backup?
2) In general, is there a (small) chance of a (failing) VM backup causing
the filesystem to be read-only afterwards? I read that qemu-guest-agent
calls fsfreeze during Proxmox VM backup. What are the impacts of fsfreeze?
How are errors handled? Are there other potential risks?

Here's the VM config:

agent: 1
bios: seabios
boot: cdn
bootdisk: scsi0
cipassword: **********
ciuser: cloudadmin
cores: 8
cpu: Skylake-Server-noTSX-IBRS
description: VM Description
hotplug: network,disk,usb
kvm: 1
machine: q35
memory: 32768
name: vm-name
net0: virtio=26:4B:29:B1:71:89,bridge=vmbr0,tag=37
numa: 0
onboot: 0
ostype: l26
protection: 1
rng0: source=/dev/urandom
scsi0: ceph:vm-108-disk-0,size=255180M,ssd=1
scsihw: virtio-scsi-single
serial0: socket
smbios1: uuid=e8736d61-9a25-4f89-9f6d-a84ef25c42cc
sockets: 1
tablet: 1
vga: qxl
vmgenid: 9d3233e9-6bc1-43e8-be50-39597b6a5034


I highly appreciate any answers and thoughts!

All the best

Benjamin