Open Stack

On 26/10/2013, at 12:01 AM, David Kranz <dkranz at redhat.com> wrote:

> On 10/25/2013 09:10 AM, Sean Dague wrote:
>> On 10/25/2013 08:39 AM, David Kranz wrote:
>>> A patch was submitted with some new tests of this api
>>> https://review.openstack.org/#/c/49778/. I gave a -1 because if a
>>> negative test to shutdown a host fails, a compute node will be shutdown.
>>> The author thinks this test should be part of tempest. My issue was that
>>> we should not have tempest tests for apis that:
>>> 
>>> 1. May corrupt the underlying system (that is part of the reason we
>>> moved whitebox out)
>> 
>> I really felt the reason we moved out whitebox is that OpenStack internals move way to fast to have them being validated by an external system. We have defined surfaces (i.e. API) and that should be the focus.
> It was also because we were side-effecting the database out-of-band.
>> 
>>> 2. Can have only negative tests because positive ones could prevent
>>> other tests from executing
>> 
>> Honestly, trying to shut down the host with invalid credentials seems like a fair test. Because if we fail, we're going to know really quick when everything blackholes.
>> 
>> In the gate this is slightly funnier because tempest is running on the same place as the host, however it seems like a sane check to have in there.
>> 
>>    -Sean
> OK, I don't feel strongly about it. Just seemed like a potential landmine.
> 

I think this is something we want to test in the gate. But perhaps there could be a tag for these sorts of test cases that some people may not want to risk running on their system so they can exclude them easily?

Chris