cinder-volume loss amqp connection when deleting of amount of cinder-volumes

[Leontii Istomin]

Detailed bug description:
 cinder-volume loss amqp connection when deleting of amount of cinder-volumes.
 from cinder-volume log: http://paste.openstack.org/show/520894/
Steps to reproduce:
 1. deploy Fuel-9.0-mos-495
 2. deploy cluster
 4. create 2000 volumes
 5. delete 2000 volumes
Expected results:
 all cinder volumes deleted
Actual result:
 Cinder volumes are in "deleting" status
Reproducibility:
 each time
Workaround:
 restart cinder-volume services
Impact:
 OpenStack cluster operating
Description of the environment:
- Operation system: ubuntu
- Versions of components: MOS-9.0
- Reference architecture: 3 controllers, 176 computes, 20 computes+Ceph, Ceph-for-all
- Network model: vxlan+dvr
- Related projects installed: LMA (disabled collectors and collectd on controller nodes)
Additional information:
  cinder-volume.log from one of controllers is attached