QMP Backup Timeout on all VMs – LXC Works but Shows PBSPlugin Warning

L

Lukas Petersen

New Member
Nov 10, 2023
10
2
3
Hi everyone,

I am running into a strange issue with my VM backups and could really use some help.

Environment:
- Proxmox VE 9.0.10
- Proxmox Backup Server 3.4.2
- Storage is configured as "PiBackup" and resolves correctly via DNS (pi-backup.lan)

What happens:
- All LXC containers back up successfully (but show a warning, see below)
- All VMs fail – no matter which one I try

Example log from a failing VM:
-------------------------------------------------
INFO: starting new backup job: vzdump 113 --remove 0 --node proxmox2 --notification-mode notification-system --mode snapshot --storage PiBackup --notes-template '{{guestname}}'
INFO: Starting Backup of VM 113 (qemu)
INFO: Backup started at 2025-09-28 19:24:46
INFO: status = running
INFO: VM Name: haos15.1
INFO: include disk 'scsi0' 'loopzfs:vm-113-disk-0' 32G
INFO: include disk 'efidisk0' 'loopzfs:vm-113-disk-1' 528K
INFO: backup mode: snapshot
INFO: ionice priority: 7
INFO: creating Proxmox Backup Server archive 'vm/113/2025-09-28T17:24:46Z'
ERROR: VM 113 qmp command 'backup' failed - got timeout
INFO: aborting backup job
INFO: resuming VM again
ERROR: Backup of VM 113 failed - VM 113 qmp command 'backup' failed - got timeout
INFO: Backup job finished with errors
-------------------------------------------------

On the PBS side (`journalctl -u proxmox-backup-proxy -f`) I see no log entries at all while the job runs – it's as if the request never even reaches the PBS.

Additional information:
- `proxmox-backup-client status --repository pi-backup.lan:HDD1` works and shows storage usage correctly
- DNS resolution works (`ping pi-backup.lan`)
- Qemu Guest Agent is disabled (enabling it makes no difference)
- Tried multiple VMs, all show the same behavior

LXC backup warning (but they still complete):
-------------------------------------------------
Wide character in setenv at /usr/share/perl5/PVE/Storage/PBSPlugin.pm line 330.
-------------------------------------------------

It feels like QEMU hangs during the backup start phase before any data is sent to PBS.

Has anyone seen this combination of `qmp command 'backup' timeout` + `PBSPlugin.pm wide character warning` before?
Any pointers what to check next would be appreciated.
 
Hi,
does the issue still occur after you upgrade to the latest version of Proxmox Backup Server 3.4? Does backing up a newly created VM without any disks work? Can you still try to add a second entry of the storage using the IP instead of hostname and see if you can back up to that storage?
 
Thanks for the suggestions!

1. I tried adding a second storage entry using the IP address instead of the hostname – unfortunately, this does not change anything. The same timeout occurs when backing up VMs.

2. I also tested backing up a VM without any disks attached.
Here is the log:

INFO: starting new backup job: vzdump 106 --notes-template '{{guestname}}' --storage PBS-Synology --node proxmox2 --notification-mode notification-system --mode snapshot --remove 0
INFO: Starting Backup of VM 106 (qemu)
INFO: Backup started at 2025-09-29 12:08:01
INFO: status = running
INFO: VM Name: TestBackup
INFO: backup mode: snapshot
INFO: ionice priority: 7
INFO: skip unused drive 'loopzfs:vm-106-disk-0' (not included into backup)
INFO: creating Proxmox Backup Server archive 'vm/106/2025-09-29T10:08:01Z'
INFO: backup contains no disks
Wide character in setenv at /usr/share/perl5/PVE/VZDump/QemuServer.pm line 740.
INFO: starting diskless backup
INFO: /usr/bin/proxmox-backup-client backup --repository root@pam@my-example-domain.net:Backup-Synology --backup-type vm --backup-id 106 --backup-time 1759140481 qemu-server.conf:/var/tmp/vzdumptmp2229566_106/qemu-server.conf
INFO: Starting backup: vm/106/2025-09-29T10:08:01Z
INFO: Client name: proxmox2
INFO: Starting backup protocol: Mon Sep 29 12:08:01 2025
INFO: Downloading previous manifest (Tue Aug 12 03:18:10 2025)
INFO: Upload config file '/var/tmp/vzdumptmp2229566_106/qemu-server.conf' to 'root@pam@my-example-domain.net:8007:Backup-Synology' as qemu-server.conf.blob
INFO: Duration: 1.05s
INFO: End Time: Mon Sep 29 12:08:02 2025
INFO: adding notes to backup
Wide character in setenv at /usr/share/perl5/PVE/Storage/PBSPlugin.pm line 330.
INFO: Finished Backup of VM 106 (00:00:01)
INFO: Backup finished at 2025-09-29 12:08:02
Wide character in setenv at /usr/share/perl5/PVE/Storage/PBSPlugin.pm line 330.
INFO: Backup job finished successfully
INFO: notified via target `mail-to-root`
TASK OK

So the diskless VM backup does work (despite the "Wide character in setenv" warnings),
but VMs with disks still fail.

As mentioned, all of my CT backups work fine.
I can also confirm that the same behavior occurs on a second PBS running version 4.0.15 –
so this seems not to be version-specific.
 
The diskless backup is also done via proxmox-backup-client directly, while a VM backup with disks is done via a QEMU library for PBS. So the issue is most likely related to that library. Can you try starting the backup in a clean environment? Could you share the output of env (censoring anything sensitive in a way that one can still guess what the original structure of the data was)?
 
Here is the output of env on my PVE node (sensitive data removed but structure kept):

SHELL=/bin/bash
PWD=/root
LOGNAME=root
XDG_SESSION_TYPE=tty
MOTD_SHOWN=pam
HOME=/root
LANG=en_US.UTF-8
SSH_CONNECTION=<redacted>
XDG_SESSION_CLASS=user
TERM=xterm-256color
USER=root
SHLVL=1
XDG_SESSION_ID=6068
XDG_RUNTIME_DIR=/run/user/0
SSH_CLIENT=<redacted>
LC_ALL=C
PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
SSH_TTY=/dev/pts/0
_=/usr/bin/env

I also tried running the backup in a completely clean environment with:

"env -i PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin LANG=en_US.UTF-8 vzdump 113 --storage PiBackup --mode snapshot --remove 0"

Unfortunately, the problem still occurs in this clean environment.
 
Does your PBS password maybe contain a non-ASCII character? Could you share the entry for PBS in the /etc/pve/storage.cfg? Is the MTU correctly configured for the network?
 
Thanks for the suggestions! I was finally able to solve the problem.
The root cause was that my PBS password contained a non-ASCII character.

After changing the PBS password to use only ASCII characters, the VM backups started working normally again
 
  • Like
Reactions: fiona
Hmm, it should be possible to use non-ASCII characters since libpve-storage-perl >= 9.0.9 since this commit. Could it be that the password was set before that? You can check /var/log/apt/history.log.

EDIT: apparently that commit did not fix the case when using the QEMU library for PBS. I can reproduce the issue and will try to come up with a fix!
 
Last edited:
  • Like
Reactions: Lukas Petersen
Thanks a lot for the quick investigation!
Good to hear that patches have already been sent to the development mailing list.


For now, I’ll stick with the ASCII-only password workaround, but I’ll happily test again once the fix is released.


Really appreciate the support!!
 
You must log in or register to reply here.
Top Bottom