Open Stack

On 3/5/2016 9:48 AM, Adam Young wrote:
> On 03/05/2016 12:27 AM, Chris Friesen wrote:
>> On 03/04/2016 03:42 PM, Matt Riedemann wrote:
>>>
>>>
>>> On 3/3/2016 9:14 PM, Zhenyu Zheng wrote:
>>>> Hm, I found out the reason:
>>>> https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/servers.py#L1139-L1145
>>>>
>>>>
>>>> here we filtered out parameters like "deleted", and that's why the API
>>>> behavior is like above mentioned.
>>>>
>>>> So should we simple add "deleted" to the tuple or a microversion is
>>>> needed?
>>
>>> Nice find. This is basically the same as the ip6 case which required
>>> microversion 2.5 [1]. So I think this is going to require a
>>> microversion in
>>> Newton to fix it (which means a blueprint and a spec since it's an
>>> API change).
>>> But it's pretty trivial, the paperwork is the majority of the work.
>>>
>>> [1] https://review.openstack.org/#/c/179569/
>>
>> Does it really need a spec given that microversions are documented in
>> the codebase?
>>
>> That almost seems like paperwork for the sake of following the rules
>> rather than to serve any useful purpose.
>>
>> Is anyone really going to argue about the details?
>>
>
> I've been lurking on this discussion. I was worried that you were going
> to try to hard code "admin" somewhere in here.
>
> If the only change is that the currently accepted set of params is
> enforced with the current set of policy rules, just in a slightly
> different place, it will not break people, and thus I would think no
> microversion is essential.  However, if the the user might need to test
> which way policy is enforced in order to use the right code path when
> doing a client call, then a microversion would be needed.
>
>
>
>> Chris
>>
>>
>> __________________________________________________________________________
>>
>> 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
>

The ip6 case and microversion 2.5 is exactly the same scenario and sets 
precedent here, so yes we need a microversion which makes it an API 
change which according to our policy requires a spec. I realize it's 
trivial, but them's the rules.

As far as I can tell, this is latent behavior since forever and no one 
has freaked out about it before, so I don't think doing things by the 
book and doing that in Newton is going to cause any problems.

-- 

Thanks,

Matt Riedemann