Servers not evacuated when a node goes down

[Willie Nelligan]

MOS 5.1: VMs not evacuated when blade goes down
Time of test ~9:26 on 2014-10-23

1.2. Detailed trouble description
Node-7 is rebooted with two VMs defined on it.

root@node-3:~# nova list --fields name,status,task_state,host,tenant_id,Networks --all-tenants
+--------------------------------------+--------------------------------+--------+------------+--------+----------------------------------+--------------------+
| ID | Name | Status | Task State | Host | Tenant Id | Networks |
+--------------------------------------+--------------------------------+--------+------------+--------+----------------------------------+--------------------+
| cc70be0e-fc5e-4f4f-b260-01a055dff09c | MOS51 | ACTIVE | None | node-7 | be18d04cae454f3db0f73fc3047d51c3 | private=10.0.10.2 |
| 5ea8430b-1bd8-4670-9b38-f261c113b0ae | MOS51_2 | ACTIVE | None | node-6 | be18d04cae454f3db0f73fc3047d51c3 | private=10.0.10.4 |
| d2bc74c9-c73a-4b2f-9eab-2f8a0be6495d | MOS51_2 | ACTIVE | None | node-7 | be18d04cae454f3db0f73fc3047d51c3 | private=10.0.10.5 |
| 1e39805e-c10d-4809-bafb-9903ae35f003 | MOS51_3 | ACTIVE | None | node-6 | be18d04cae454f3db0f73fc3047d51c3 | private=10.0.10.6 |
| bf2451a1-7b26-47a2-98bc-e2ccdcd74e42 | OST.node-1---20141024-071805-0 | ACTIVE | None | node-5 | be18d04cae454f3db0f73fc3047d51c3 | OST_net=99.99.0.14 |
+--------------------------------------+--------------------------------+--------+------------+--------+----------------------------------+--------------------+
root@node-3:~#

The VMs on node-7 are never evacuated when the reboot of the blade occurs.

Node-7 recovers and "virsh list" shows that the VMs are still present:
root@node-7:~#
root@node-7:~# virsh list
 Id Name State
----------------------------------------------------
 2 instance-00000001 running
 3 instance-00000007 running

root@node-7:~#

Even with the evacuation parameter set in the VMs metadata, there is no change; evacuation is not working.

[Willie Nelligan (willie-nelligan)]

Configuration data
HP C7000 node with MOS5.1 installed by fuel:

[root@fuel ~]# fuel node list
id | status | name | cluster | ip | mac | roles | pending_roles | online
---|--------|------------------|---------|---------------|-------------------|---------------------------|---------------|-------
6 | ready | Untitled (00:18) | 1 | 192.168.3.105 | 72:f0:a1:89:ae:4b | compute | | True
1 | ready | Untitled (00:04) | 1 | 192.168.3.100 | 4e:90:fb:ee:ef:4d | cinder, controller, mongo | | True
5 | ready | Untitled (00:14) | 1 | 192.168.3.104 | b2:1a:8f:ad:d6:4f | compute | | True
2 | ready | Untitled (00:08) | 1 | 192.168.3.101 | 72:f5:c6:c7:12:48 | cinder, controller, mongo | | True
4 | ready | Untitled (00:10) | 1 | 192.168.3.103 | 2a:1b:0b:7c:6f:45 | compute | | True
3 | ready | Untitled (00:0c) | 1 | 192.168.3.102 | 22:63:ec:e1:b0:40 | cinder, controller, mongo | | True
7 | ready | Untitled (00:1c) | 1 | 192.168.3.106 | ae:59:66:7b:31:48 | compute | | True
[root@fuel ~]#

[Dmitry Mescheryakov (dmitrymex)]

Willie, AFAIK the auto-evacuation feature is not implemented yet in OpenStack. If it is implemented, please point to the documentation or piece of code implementing it. Otherwise, I will move the bug to invalid state.

[Roman Podoliaka (rpodolyaka)]

The best we can do is Mitaka, so I'm marking this as Won't Fix for 8.0