Open Stack

On 05/21/2013 11:43 PM, Clark Boylan wrote:
> On Tue, May 21, 2013 at 8:27 PM, Clark Boylan <clark.boylan at gmail.com> wrote:
>> On Tue, May 21, 2013 at 7:50 PM, Sandy Walsh <sandy.walsh at rackspace.com> wrote:
>>> I was close to getting my branch landed when a rebase brought the testr
>>> changes in ... https://review.openstack.org/#/c/29047/
>>>
>>> Now, my tests are failing in an odd way (and they used to pass).
>>>
>> snip
>>> But when I apply the 'raise e' patch as suggested, it shows the
>>> underlying issue as being:
>>> http://paste.openstack.org/show/37563/
>>>
>>> Which might suggest the migrations aren't being set up correctly?
>>> (odd as the other sqlalchemy tests are passing)
>>>
>>> I'm sort of stumped here ... suggestions?
>>>
>>> -S
>>>
>>> My branch, if you care to try for yourself ...
>>> https://review.openstack.org/#/c/29112/11
>>>
>> According to Jenkins' comment on this change the unittests ran
>> successfully. And the test called out in paste 37563 runs successfully
>> according to the subunit log. Where/How are the tests being run when
>> they fail?
>>
> I suppose it is also worth mentioning that testr may run tests in
> different orders (and different processes if the tests are run in
> parallel but ceilometer does not appear to do that yet). It may help
> others debug if you share a copy of your subunit log. Would be located
> at .testrepository/X where X is the test ID reported by testr at the
> end of testing. And finally you can run testr's test bisection tool to
> try and track down potential inter test conflicts by running `testr
> run --analyze-isolation` after a failed test run (testr uses the
> subunit log of the previous test run to determine what order tests
> were run in when they failed).

When doing the nova transition Clark and I also found that artificially 
increasing the parallelism in testr (like going 64 way on a 48GB memory 
box) made the test state leaks fall out a lot faster.

	-Sean

-- 
Sean Dague
http://dague.net