Should rollback the instance if finish_resize fails

[Charlotte Han]

Resize failed in finish_resize function, the instance disappeared and can not rollback.

log is :
http://paste.openstack.org/show/137968/

[Madhurya (madhurya-jesu)]

Hi Rong Han, can you please elaborate your bug. When I changed the '-allow_resize_to_same_host=true' and '-allow_migrate_to_same_host=true' in nova.conf and restarted nova-compute and nova-scheduler, I was able to resize the instance. So can please post your nova.conf file.

[Charlotte Han (hanrong)]

# Allow destination machine to match source for resize. Useful
# when testing in single-host environments. (boolean value)
#allow_resize_to_same_host=false

[Charlotte Han (hanrong)]

I use the default configuration of 'allow_resize_to_same_host=false' in nova.conf.

[Robert Collins (lifeless)]

So it sounds like the default configuration will break vms on resize?

[Jian Wen (wenjianhn)]

Can anyone else reproduce this bug?

[Charlotte Han (hanrong)]

Before resize instance, stop neutron-openvswitch-agent, then resize/migrate instance, the instance is error.

[Jian Wen (wenjianhn)]

I reproduced the bug by adding an exception to finish_migration().
" vif_type=binding_failed" means neutron failed to bind the port to the compute host.
You cannot rollback the instance since Neutron doesn't support so.
I don't think we can fix the bug in Nova.

p.s.
The error log is not friendly to users.

[Charlotte Han (hanrong)]

if destination host's neutron server is unavailable, we should not resize to this destination host, or we can let it revert to it's source host.

[Jian Wen (wenjianhn)]

You are right.

Looks like we are able to revert it manually now.
See bug 1296519.

On Thu, Nov 19, 2015 at 3:39 PM, Rong Han ZTE <email address hidden> wrote:

> if destination host's neutron server is unavailable, we should not
> resize to this destination host, or we can let it revert to it's source
> host.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1396003
>
> Title:
> Resize failed in finish_resize function, the instance disappeared and
> can not rollback.
>
> Status in OpenStack Compute (nova):
> Invalid
>
> Bug description:
> Resize failed in finish_resize function, the instance disappeared and
> can not rollback.
>
> log is :
> http://paste.openstack.org/show/137968/
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/nova/+bug/1396003/+subscriptions
>

--
Best,

Jian

[Charlotte Han (hanrong)]

I try the same operation with kilo version, the same error is occurring.

[Charlotte Han (hanrong)]

and I reset-state --active, then hard reboot this instance, the instance is error.
log is as follow:
http://paste.openstack.org/show/479397/

[Charlotte Han (hanrong)]

I try the same operation with Mikata version, the same error is occurring.

see https://bugs.launchpad.net/nova/+bug/1586309, delete operation could not clear source compute node resource.

The openvswitch agent service on Destination compute node is active, but public net is not available for some reason I don't know on destination node.

Devstack openstack environment, two networks were created, I had two compute nodes.
When I created instance using public net, resize failed with Unexpected vif_type.
When I created instance using private net, resize successfully.
So I think when I chose some network which can not resize to another compute node successfully, please let this instance rollback to source compute node.
[stack@SBCJSlot5Rack2Centos7 instances]$ nova net-list
+--------------------------------------+---------+------+
| ID | Label | CIDR |
+--------------------------------------+---------+------+
| 44652d4f-95ee-4785-8c16-de74bbb722c0 | public | None |
| d40836c3-f036-4151-bbaa-cf92125ce552 | private | None |

[Charlotte Han (hanrong)]

Can we use "revert_resize" to recover the error state of the instances after they had been migrated/resized failed?

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/334747

[Sean Dague (sdague)]

Patch in merge conflict

[OpenStack Infra (hudson-openstack)]

Change abandoned by Sean Dague (<email address hidden>) on branch: master
Review: https://review.openstack.org/334747
Reason: This review is > 6 weeks without comment, and failed Jenkins the last time it was checked. We are abandoning this for now. Feel free to reactivate the review by pressing the restore button and leaving a 'recheck' comment to get fresh test results.