Comment 2 for bug 1254872

What this error message indicates is that two separate threads have made libvirt API calls against the same VM. One of the calls has either hung completely, or is taking a very long time to respond, causing the second API call to report this error message.

Here we're seeing the error message in response to virDomainAttachDeviceFlags. Unfortunately we need to really see what
the original API call was which is hard to find out.

With some libvirt versions there have been bugs which caused this error message, so its important to know what libvirt version shows this error. Indeed if the gate runs on hosts with a variety of libvirt / qemu versions, it'd be useful to figure out if it only occurs on hosts with specific versions of libvirt/qemu.

To diagnose this further likely requires turning up the libvirt logging level.

For Nova I'd suggest setting

  export LIBVIRT_LOG_FILTERS="1:libvirt"
  export LIBVIRT_LOG_OUTPUTS=1:file:/var/log/libvirt/nova-cpu.log"

before starting the nova-cpu processes. This captures client side traces

It might also be useful to enable logging on the libvirtd server side by editing /etc/libvirt/libvirtd.log and settings

  log_filters="1:libvirt 1:qemu"
  log_outputs="1:file:/var/log/libvirt/libvirtd.log"

this captures server side traces.