Guest fails to start / immediately in paused state
Software environment
- Operating system: CentOS 8
- Architecture: x86_64
- kernel version: 4.18.0-305.3.1.el8.x86_64
- libvirt version: 7.3.0
- Hypervisor and version: kvm/qemu 6.0.0
- getenforce: Permissive
Description of problem
A guest fails to start on the system. It immediately falls into paused state. This happens for both, starting the guest or live migrating it onto the host.
In order to remove the guest from the paused state a restart of the libvirtd daemon is necessary before the two commands
virsh destroy ae01fff2-833f-4196-aea1-53aaaccd1156
virsh undefine ae01fff2-833f-4196-aea1-53aaaccd1156
finish successfully.
The guest can be booted without any problems on libvirt 7.1.0 / qemu 5.2 / nearly vanilla 5.10.29 Linux
Steps to reproduce
Not sure. I can reproduce reliably with given guest. However, using a similar (almost identical domain XML) produces no bad results.
Additional information
Domain XML
A nearly identical domain XML works without issues:
The one that does not work is only different in a few minor details (I think).
Debug log of the daemon
The domain UUID of the problematic guest is ae01fff2-833f-4196-aea1-53aaaccd1156
. The guest was started on the host at around 10:51
After the domain failed to start the libvirtd daemon was restarted and the domain was destroyed and undefined. (Also in the logs.)