backup stop mode doesn't stop the VM

[ad-computing]

Hello,

I don't know exactly when the problem started, but the backup in "stop mode" do not stop actualy the VM. I'm currently running PVE 9.1.6.
This problem occurs with or without qemu-guest-agent active.
The VM are mostly running DEBIAN 13. but It's the same problem with a Windows server 2022.

 

[driley]

Hi,

the backup "stop mode" does not forcefully stop the VM. It tries to perform a clean shutdown sequence.
The guest os might be ignoring the shutdown request or it might be blocked by some process.

Have you tried shutting down the VM manually using the "shutdown" command? If so did this work?

What is the exact error message you get when performing the backup?

You can also try adjusting the stop timeout in '/etc/vzdump.conf' by uncommeting the line 'stopwait' and setting a higher value then the default (10).
for example:
stopwait: 20

 

[ad-computing]

Thank's for your quick answer.

I've no error message :
NFO: starting new backup job: vzdump 201 --remove 0 --compress zstd --storage backup --mode stop --node pmx1 --notification-mode notification-system --notes-template '{{guestname}}'
INFO: Starting Backup of VM 201 (qemu)
INFO: Backup started at 2026-03-11 16:50:39
INFO: status = running
INFO: backup mode: stop
INFO: ionice priority: 7
INFO: VM Name: pmx1-bunkerweb
INFO: include disk 'scsi0' 'local:201/vm-201-disk-0.qcow2' 40G
INFO: stopping virtual guest
INFO: creating vzdump archive '/mnt/pve/backup/dump/vzdump-qemu-201-2026_03_11-16_50_38.vma.zst'
INFO: starting kvm to execute backup task
_________________________________________________________________________________________________________________
The stop task seems to work but it does nothing..

If I stop the vm from the host with the command "stop [vmid]" it works :

root@pmx1:~# qm status 201
status: running
root@pmx1:~# qm stop 201
root@pmx1:~# qm status 201
status: stopped

I modified the "/etc/vzdump.conf" stopwait with value "1" (the parameter is not in seconds but in in minutes) but it doesn’t work either
From my point of view it means that the planned backups with the stop mode are not reliables. I will script the backups between a « qm stop » and a « qm start ».

 

[driley]

Hi,

thanks for the detailed logs.

What happens if you try:
qm shutdown 201
How long does that take?

When the qemu-guest-agent is active do you get an error response when running:
qm agent 201 ping

When you deactivated the qemu-guest-agent did you make sure to uncheck the "QEMU guest agent" checkbox as well in the VMs option tab?

If none of this helps to find a solution the scripting will work for sure.

 

[ad-computing]

Hi,

Yes the QEMU Guest Agent is unchecked

root@pmx1:~# qm agent 201 ping
QEMU guest agent is not running

root@pmx1:~# time qm shutdown 201

real 0m3.689s
user 0m0.605s
sys 0m0.078s

root@pmx1:~# qm status 201
status: stopped

root@pmx1:~# time qm start 201

real 0m1.566s
user 0m0.666s
sys 0m0.165s

The script solution is only a temporary workaround, but a fix must be developped shortly..

Thanks for your support

To go further I developped a little hook, but the surprise is :

the backup starts kvm while the backup (!) :

INFO: starting kvm to execute backup task
INFO: started backup task '3e22702f-215f-4575-965a-c84c85bdbefe'
INFO: resuming VM again after 8 seconds

maybe somebody can explain exactly how the backup works ? I started to pull my hair

HOOK : #!/bin/bash #set -x echo "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++" echo " hook parameters : $*" echo "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++" if [[ "$1 $2" = "backup-start stop" ]] then VMID="$3" echo "Backup VM $VMID starts..." echo "shutdown requested" qm unlock $VMID qm shutdown $VMID echo "===> VM $VMID status :" qm status $VMID fi if [[ "$1 $2" = "pre-restart stop" ]] then VMID="$3" echo "===> VM $VMID status :" qm status "$VMID" fi if [[ "$1 $2" = "post-restart stop" ]] then VMID="$3" echo "===> VM $VMID status :" qm status "$VMID" fi if [[ "$1 $2" = "backup-end stop" ]] then VMID="$3" echo "Backup VM $VMID ended ..." qm status $VMID fi if [[ "$1" == "job-end" ]] then echo "Job finished." fi echo "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"

 

[VictorSTS]

You need to understand how backup works. Specifically in stop mode:

• VM is running
• PVE tries to do a shutdown of the VM:
  • If QEMU Agent is configured and running, shutdown signal is issued with it.
  • Else, sends an ACPI signal which guest OS may or may not honor.
• It waits up to 10 minutes for the shutdown to finish, so make sure that:
  • Your guest OS and apps are able to close within that time.
  • Your VM honors shutdown signals.
• QEMU starts the VM in backup mode and creates a snapshot of every disk at QEMU level (doesn't depend on snapshot support at storage level)
• Once the snapshot is taken (usually few seconds) two actions start almost at the same time:
  • The VM is started in normal mode (the VM is not off during the whole backup).
  • The backup process copies to backup from the snapshot.

In your log, it states clearly that he shutdown is issued to the VM:

 INFO: stopping virtual guest

root@pmx1:~# qm status 201
status: running
root@pmx1:~# qm stop 201
root@pmx1:~# qm status 201
status: stopped

Careful with this: you are issuing a stop, which kills the running VM. Use qm shutdown to tell the guest OS to shutdown cleanly.

I strongly recommend to use QEMU Agent which, among other things, will enable clean shutdowns and properly quiescence the OS (and even integrate with Windows' VSS) on snapshot mode.

Why are you using stop mode?

 

[ad-computing]

You need to understand how backup works. Specifically in stop mode:

• VM is running
• PVE tries to do a shutdown of the VM:
  • If QEMU Agent is configured and running, shutdown signal is issued with it.
  • Else, sends an ACPI signal which guest OS may or may not honor.
• It waits up to 10 minutes for the shutdown to finish, so make sure that:
  • Your guest OS and apps are able to close within that time.
  • Your VM honors shutdown signals.
• QEMU starts the VM in backup mode and creates a snapshot of every disk at QEMU level (doesn't depend on snapshot support at storage level)
• Once the snapshot is taken (usually few seconds) two actions start almost at the same time:
  • The VM is started in normal mode (the VM is not off during the whole backup).
  • The backup process copies to backup from the snapshot.

In your log, it states clearly that he shutdown is issued to the VM:

 INFO: stopping virtual guest

Careful with this: you are issuing a stop, which kills the running VM. Use qm shutdown to tell the guest OS to shutdown cleanly.

I strongly recommend to use QEMU Agent which, among other things, will enable clean shutdowns and properly quiescence the OS (and even integrate with Windows' VSS) on snapshot mode.

Why are you using stop mode?

Thank you for the explanation.

I assumed the "stop mode" in vzdump was a "clean stop", allowing a "cold backup" during a night window.

What I don't understand is why by specifying "--mode stop" it is still possible to connect to the VM being backed up.
For me the “stop” option allowed me to make a cold backup of the VM. From what you say I see that this is
false: in fact saving in "stop mode" is finaly equivalent to saving in "snapshot" mode, this was confusing for me.
Th reason why I wanted to use the "stop mode" is that we had some failure during some restore from a "snapshot mode"
so I lost my faithfull in this mode".

 

[fabian]

stop mode backup will just stop the VM and then start it paused to ensure consistency. it will then setup the copy-before-write-filters before starting actual guest execution.

 

[ad-computing]

Thanks for those explanations, the question is clear now.
I deployed in datacenters my architecture based on proxmox VE since 2019 in production and development environnement on clusters and single nodes and I’m still happy with it