Live migration doesn't work properly for Windows VM
Affects | Status | Importance | Assigned to | Milestone | ||
---|---|---|---|---|---|---|
Mirantis OpenStack | Status tracked in 10.0.x | |||||
10.0.x |
Invalid
|
High
|
MOS Nova | |||
8.0.x |
Invalid
|
High
|
Oleksandr Liemieshko | |||
9.x |
Invalid
|
High
|
MOS Maintenance |
Bug Description
When attempting a live-migration of an instance(Windows) from one compute node to another I got a situation when the instance failed to migrate properly.
Instance has "SHUTOFF" status in nova, it is in "shut off" state on an old compute node and "running" on a new one.
Moreover if you open VNC connection to the instance before migration it will continue to work in spite of the states which were provided above. But if you try to open new VNC connection to instance you will get "Failed to connect to server (code: 1006)" error.
In nova:
root@node-7:~# nova show dcd2d83c-
+------
| Property | Value |
+------
| OS-DCF:diskConfig | AUTO |
| OS-EXT-
| OS-EXT-
| OS-EXT-
| OS-EXT-
| OS-EXT-
| OS-EXT-
| OS-EXT-STS:vm_state | stopped |
| OS-SRV-
| OS-SRV-
| accessIPv4 | |
| accessIPv6 | |
| admin_internal_net network | 192.168.111.12 |
| config_drive | |
| created | 2017-01-
| flavor | m1.medium (3) |
| hostId | ce0eff158351c1c
| id | dcd2d83c-
| image | win2012r2 (7d7f4c95-
| key_name | - |
| metadata | {} |
| name | orig |
| os-extended-
| security_groups | default |
| status | SHUTOFF |
| tenant_id | 3bdaac4958d248d
| updated | 2017-01-
| user_id | 6a8a2c38b7144e6
+------
On an old compute node:
root@node-9:~# virsh list --all | grep "instance-0000000c"
- instance-0000000c shut off
On a new compute node:
root@node-8:~# virsh list --all | grep "instance-0000000c"
61 instance-0000000c running
After "Hard Reboot Instance" it is "running" on both compute nodes
root@node-8:~# date
Thu Jan 19 11:07:19 UTC 2017
root@node-8:~# virsh list --all | grep "instance-0000000c"
61 instance-0000000c running
root@node-9:~# date
Thu Jan 19 11:07:13 UTC 2017
root@node-9:~# virsh list --all | grep "instance-0000000c"
142 instance-0000000c running
root@node-7:~# nova show dcd2d83c-
+------
| Property | Value |
+------
| OS-DCF:diskConfig | AUTO |
| OS-EXT-
| OS-EXT-
| OS-EXT-
| OS-EXT-
| OS-EXT-
| OS-EXT-
| OS-EXT-STS:vm_state | active |
| OS-SRV-
| OS-SRV-
| accessIPv4 | |
| accessIPv6 | |
| admin_internal_net network | 192.168.111.12 |
| config_drive | |
| created | 2017-01-
| flavor | m1.medium (3) |
| hostId | ce0eff158351c1c
| id | dcd2d83c-
| image | win2012r2 (7d7f4c95-
| key_name | - |
| metadata | {} |
| name | orig |
| os-extended-
| progress | 0 |
| security_groups | default |
| status | ACTIVE |
| tenant_id | 3bdaac4958d248d
| updated | 2017-01-
| user_id | 6a8a2c38b7144e6
+------
Steps to reproduce:
- Fuel 8.0
- 1 controller, 2 computes
- VLAN
- Ceph for all
- image for Windows (https:/
Scenario:
1. Create new VW from Windows image
2. Open VNC connection to the instance before migration
3. Try to migrate "Live Migrate Instance" few times
Changed in mos: | |
importance: | Undecided → High |
assignee: | nobody → MOS Nova (mos-nova) |
tags: | added: cus |
tags: |
added: customer-found removed: cus |
tags: | added: support |
tags: | added: area-nova |
Alexander, could you please provide nova logs from both compute nodes.