[Users] affects of changing compatibility setting on running cluster

[Itamar Heim]

On 03/12/2014 10:45 PM, ybronhei wrote:
> On 03/12/2014 04:33 PM, Itamar Heim wrote:
>> On 03/12/2014 04:32 PM, ybronhei wrote:
>>> On 03/12/2014 04:25 PM, Dan Kenigsberg wrote:
>>>> On Wed, Mar 12, 2014 at 10:15:21AM +0200, Itamar Heim wrote:
>>>>> On 03/12/2014 10:11 AM, Sven Kieske wrote:
>>>>>>
>>>>>>
>>>>>> Am 11.03.2014 17:38, schrieb Itamar Heim:
>>>>>>>
>>>>>>> i understood this to "you can't use 4.14 with 3.4.0".
>>>>>>
>>>>>> Well I examined BZ 1067096 again, this is what did not work:
>>>>>>
>>>>>> Steps to Reproduce:
>>>>>> 1.  Run Engine 3.3 Compatibility level 3.2 or 3.3
>>>>>> 2.  Add node (vdsm 4.14)
>>>>>>
>>>>>> Actual results:
>>>>>> Host is installed with VDSM version (4.14) and cannot join cluster
>>>>>> which
>>>>>> is compatible with VDSM versions [4.13, 4.9, 4.11, 4.12, 4.10].
>>>>>>
>>>>>> Pay attention to the engine version, it states 3.3. not 3.4.0
>>>>>>
>>>>>> So my conclusion is, you can't install vdsm 4.14. when you want
>>>>>> to use engine 3.3.
>>>>>>
>>>>>> Am I reading something wrong?
>>>>>>
>>>>>> Here's the link again:
>>>>>> https://bugzilla.redhat.com/show_bug.cgi?id=1067096#c0
>>>>>>
>>>>>
>>>>> this can be fixed via engine config, but is not supposed to be
>>>>> needed, as vdsm is supposed to have
>>>>> vdsm/dsaversion.py.in:    'supportedENGINEs': ['3.0', '3.1', '3.2',
>>>>> '3.3', '3.4'],
>>>>>
>>>>> danken/eli - thoughts?
>>>>
>>>> I do not really understand why we have this recurrent Engine bug, of
>>>> ignoring supportedENGINEs; I think Yaniv does.
>>>>
>>>
>>> i recall asking for detailed explanation for why do we have 3 different
>>> restrictions , one for supportedEngines, one for the allowed vdsm
>>> versions, and one for the cluster level?
>>>
>>> if vdsm version is supported why isn't it in engine's capabilities yet?
>>
>> because its a vdsm version which was released after that engine was
>> released, hence the vdsm package has to declare its supporting the old
>> engine.
>>
>
> so its an hack to allow users to add beta vdsm version to new engine
> release?.. sounds like that

no. its to add the next release of vdsm to a previously released engine. 
but that's not supposed to be an issue, since its supposed to already 
report the previous version of engine (which it does)

>
> please reopen if it reproduced
>
>>>
>>> the issue was verified in
>>> https://bugzilla.redhat.com/show_bug.cgi?id=1016461 , please reopen if
>>> it still appears
>>>
>>>
>>>
>>
>
>