I've been working on a related problem, and have also noticed this delay.
These calls happen on separate compute hosts, and lots of things happen in
between. Although the goal of my patch isn't to address the issue you
describe, I have noticed that this patch significantly reduces it anyway:
I'd be interested to know if you see the same in your environment. As an
unmerged RPC change, obviously don't go putting this in production!
On Thu, Sep 8, 2016 at 10:21 AM, Jingting <kangjingt...@sina.com> wrote:
> Hi All:
> As ,  shows, at the last stage in VM migration, it will call
> unplug_vifs(), and then migrate_instance_finish(), in which update_port
> will be called immediately.
> I think the two steps should be executed without no disturb., but in my
> test, there is at least 3 seconds between unplug vif and update_port.
> As I want to do something with my controller at the two points, so I have
> tested it many times, and the lasting time remains about 3s.
> Any comment will be appreciate! Thanks!
>  http://paste.openstack.org/show/198298/
>  http://paste.openstack.org/show/568983/
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
Red Hat Engineering, Virtualisation Team
Phone: +442070094448 (UK)
OpenStack Development Mailing List (not for usage questions)