Evacuation vm with DPDK doesnt work on cinder

[Kristina Berezovskaia]

Detailed bug description:
 VM with DPDK has an ERROR state after evacuation on env with cinder

Steps to reproduce:
 1) Deploy env with DPDK and cinder
 2) Create new flavor:
nova flavor-create m1.small.hpgs auto 512 1 1
nova flavor-key m1.small.hpgs set hw:mem_page_size=2048
 3) Boot vm with DPDK and new flavor
cinder create --image-id $(nova image-list | grep ubuntu_image | awk '{print $2}') --name bootable-volume2 10
nova boot --flavor m1.small.hpgs --nic net-id=$(nova net-list | grep admin_internal_net | awk '{print $2}') --block_device_mapping vda=<your_id> --availability-zone nova:node-4.test.domain.local volume-backed-vm1
 4) Stop nova-compute
 5) Evacuate vm
Expectes result: vm is on proper node
Actual result: vm is in error state

Description of the environment:
cat /etc/fuel_build_id:
 358
cat /etc/fuel_build_number:
 358
cat /etc/fuel_release:
 9.0
cat /etc/fuel_openstack_version:
 mitaka-9.0
rpm -qa | egrep 'fuel|astute|network-checker|nailgun|packetary|shotgun':
 fuel-release-9.0.0-1.mos6345.noarch
 fuel-bootstrap-cli-9.0.0-1.mos282.noarch
 fuel-migrate-9.0.0-1.mos8386.noarch
 rubygem-astute-9.0.0-1.mos746.noarch
 fuel-misc-9.0.0-1.mos8386.noarch
 network-checker-9.0.0-1.mos72.x86_64
 fuel-mirror-9.0.0-1.mos136.noarch
 fuel-openstack-metadata-9.0.0-1.mos8707.noarch
 fuel-notify-9.0.0-1.mos8386.noarch
 nailgun-mcagents-9.0.0-1.mos746.noarch
 fuel-provisioning-scripts-9.0.0-1.mos8707.noarch
 python-fuelclient-9.0.0-1.mos315.noarch
 fuelmenu-9.0.0-1.mos270.noarch
 fuel-9.0.0-1.mos6345.noarch
 fuel-utils-9.0.0-1.mos8386.noarch
 fuel-setup-9.0.0-1.mos6345.noarch
 fuel-library9.0-9.0.0-1.mos8386.noarch
 shotgun-9.0.0-1.mos89.noarch
 fuel-agent-9.0.0-1.mos282.noarch
 fuel-ui-9.0.0-1.mos2697.noarch
 fuel-ostf-9.0.0-1.mos934.noarch
 python-packetary-9.0.0-1.mos136.noarch
 fuel-nailgun-9.0.0-1.mos8707.noarch
1 controller, 2 compute with dpdk, HP and cinder, 2 compute with HP and cinder
neutron+vlan, KVM

[Kristina Berezovskaia (kkuznetsova)]

I boot vm with new flavor on node-4 with DPDK. Id - aed157cb-b0dd-48de-b309-04ed2869b109.
Stop nova-compute on node-4 and start evacuating. After ERROR state in logs I can see that vn try to schedule on node-2 wothout DPDK

[Kristina Berezovskaia (kkuznetsova)]

I stop nova service on other nodes, only 2 nodes (4 and 5) are available with DPDK. Repeat the steps but have another error.
On node-5 I have "Lazy-loading 'numa_topology' on Instance uuid b8c3707b-24a6-4add-a4ac-769fcf483725"

Logs in attachment

[Roman Podoliaka (rpodolyaka)]

DPDK support is meant to be experimental in 9.0, thus, marking as Medium (effectively means won't fix in 9.0 after SCF).