My VM backup script fails while creating the snapshot.
virsh snapshot-create-as --domain machine_1 snap --diskspec vda,file=/srv/test/test-snap.qcow2 --disk-only --atomic --no-metadata --quiesce error: Requested operation is not valid: domain is already quiesced
Even after a VM reboot, the system is still quiesced and I get the same error.
I thought quiesce means FS freeze, but this makes no sense since I can still write to the FS when logged in the faulty VMs. And this would not survive a reboot, right?
Could it be a communication issue that makes the host think the GA says the machine is quiesced while it is not?
In any case, is there a command to enquire the quiesce state (apart from attempting a snapshot and see if I get an error)?
Assuming the faulty VMs went quiesced after a unreproducible error, I could fix that by exiting quiesced state, whatever that means. Is there a virsh command to unquiesce the VM?
The whole backup procedure used to work and now it fails on 2 VMs but still works on 2 others and I can't think of any relevant difference between them.
- Host is Debian Jessie with qemu-kvm 2.8+dfsg-3~bop8+1 from backports.
- Guests are Debian Stretch with qemu-guest-agent 2.8+dfsg-6+deb9u4.
(For the record, the backup script is here on GitHub. Basically, what it does is 1/ create snapshot, 2/ copy, 3/ commit snapshot.)
If I remove the
quiesce option from the snapshot command line, things work smooth. But obviously, this is not ideal.