Open Stack

On Mon, Jun 6, 2016 at 2:03 PM, Armando M. <armamig at gmail.com> wrote:

>
>
> On 6 June 2016 at 10:06, Oleg Bondarev <obondarev at mirantis.com> wrote:
>
>> Hi,
>>
>> There are cases where it would be useful to know the version of Neutron
>> (or any other project) from API, like during upgrades or in cross-project
>> communication cases.
>> For example in https://review.openstack.org/#/c/246910/ Nova needs to
>> know if Neutron sends vif-plugged event during live migration. To ensure
>> this it should be enough to know Neutron is "Newton" or higher.
>>
>> Not sure why it wasn't done before (or was it and I'm just blind?) so the
>> question to the community is what are possible issues/downsides of exposing
>> code version through the API?
>>
>
> If you are not talking about features exposed through the API (for which
> they'd have a new extension being advertised), knowing that you're running
> a specific version of the code might not guarantee that a particular
> feature is available, especially in the case where the capability is an
> implementation detail that is config tunable (evil, evil). This may also
> lead to needless coupling between the two projects, as you'd still want to
> code defensively and assume the specific behavior may or may not be there.
>

Agree, that's why we have extensions-list in the API, right?


>
> I suspect that your case is slightly different in that the lack of a
> received event may be due to an error rather than a missing capability and
> you would not be able to distinguish the difference if not optimistically
> assume lack of capability. Then you need to make a "mental" note and come
> back to the code to assume a failure two cycles down the road from when
> your code merges. Definitely not a pretty workflow without advertising the
> new feature explicitly via the API.
>

I'd not call it a feature, but a tiny behavior change for the neutron
reference implementation: if patch https://review.openstack.org/#/c/246898/
merges in Newton then we can be sure that neutron (newton or higher) with
ml2+ovs should send a particular event to nova during live migration (if it
doesn't than it a bug, but it's another topic) and nova can be sure it
should wait for this event.


>
>
>>
>> Thanks,
>> Oleg
>>
>> __________________________________________________________________________
>> OpenStack Development Mailing List (not for usage questions)
>> Unsubscribe:
>> OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20160607/7a1f399c/attachment.html>