Hi,

I have tried to troubleshoot this issue further.

- created a vm from image on local compute storage (local disk not shared between compute nodes).

- Tried to live migrate that vm, but it failed with the same error.

I have two environments, one is on wallaby and other is on xena. The above case works fine in wallaby but not working in xena. Is there something related to live block migration changed in xena that is causing trouble ?

Ammad

On Fri, Jan 21, 2022 at 4:02 PM Ammad Syed <syedammad83@gmail.com> wrote:

Hi,

I have a weird error while performing live migration of trove instances. The instance has a root disk residing on a local compute node and a second a second disk is attached from shared storage via cinder. There are two networks attached with the instance, one is vlan backed and other one is geneve.

When I try to perform its live migration, it fails. I see below errors in libvirt logs and the same in nova-compute logs.

Jan 19 15:56:54 host03.cloud.com libvirtd[731030]: Cannot start job (query, none, none) for domain instance-00000819; current job is (none, none, migration in) owned by (0 <null>, 0 <null>, 0 remoteDispatchDomainMigratePrepare3Params (flags=0x9b)) for (0s, 0s, 61s)
Jan 19 15:56:54 host03.cloud.com libvirtd[731030]: Timed out during operation: cannot acquire state change lock (held by monitor=remoteDispatchDomainMigratePrepare3Params)
Jan 19 15:36:30 host03.cloud.com libvirtd[731030]: migration successfully aborted

Any advise how to fix it ? I am using below libvirt version.

libvirt 6.0
qemu-kvm 4.2
kernel 5.4
nova 24.0

Ammad