Bug 1967715

Created attachment 1788873 [details]
libvirtd.log

Description of problem:

When a VM is migrating to another host and the VM is destroyed on the source host during the migration, some libvirt calls on the given VM start failing with 

  Timed out during operation: cannot acquire state change lock (held by monitor=remoteDispatchDomainMigratePerform3Params)

This continues until libvirtd is restarted, even if the VM is no longer present on the host. It prevents the given VM from starting on the host again.

Version-Release number of selected component (if applicable):

libvirt-daemon-kvm-7.0.0-14.module+el8.4.0+10886+79296686.x86_64
qemu-kvm-5.2.0-16.module+el8.4.0+10806+b7d97207.x86_64

How reproducible:

100% on my laptop (in RHV running in a nested virtualization).

Steps to Reproduce:
1. Migrate VM to another host.
2. While the migration is still running, run `virsh destroy' on the VM on the source host.
3. Try to start the same VM on the given host, or start it on another host and try to migrate it to the original host, or try running `virsh undefine' on the VM, even if it is not present on the host anymore.

Actual results:

The operations fail with "Timed out during operation: cannot acquire state change lock (held by monitor=remoteDispatchDomainMigratePerform3Params)".

Expected results:

The VM can be wiped out from the host (if needed) and started there normally again.

Additional info:

Attaching libvirtd.log demonstrating the problem.

A very similar problem has been observed in BZ 1966121 when migrating VMs (without destroying them during the migrations) on POWER. This one is not easily reproducible and has been observed only about twice. From the logs, it looks the same, but I don't have any idea whether it can have the same internal cause. Please see BZ 1966121 for more information and logs.