Open Stack

You can pull the subunit log and use subunit-ls to see the tests that
ran, and do a testr list-tests to get a set of all the tests: the one
that bailed out must be one of the runs that doesn't report having
been run [or possibly the very last one that ran in a worker - but
lets assume not]. So subtract the ones that ran from the ones that
should have run; then run testr run --load-list $YOURLISTOFTESTS. One
or more tests will run and it will stop again. Repeat removing the
tests that ran (using testr last --subunit | subunit1to2 | subunit-ls)
and running until its bailing out immediately (due to testtools having
a deterministic test order [in the absence of optimisers or
randomisers], it will always be running in the same order. Once its
exiting immediately, its probably the first test in the set that you
ran, and you can see the order testtools was going to run via 'testr
list-tests'. If it doesn't exit immediately at any point and you end
up with an empty set, then its probably a two-test interaction - but
they don't tend to exit hard IME, so lets cross that bridge if you hit
it.

-Rob



On 12 February 2014 16:49, Gary Duan <garyduan at gmail.com> wrote:
> Clark,
>
> You are right. The test must have been bailed out. The question is what I
> should look for. Even a successful case has a lot of Traceback, ERROR log in
> subunit_log.txt.
>
> Thanks,
> Gary
>
>
> On Tue, Feb 11, 2014 at 7:19 PM, Clark Boylan <clark.boylan at gmail.com>
> wrote:
>>
>> On Tue, Feb 11, 2014 at 7:05 PM, Gary Duan <garyduan at gmail.com> wrote:
>> > Hi, Clark,
>> >
>> > Thanks for your reply.
>> >
>> > I thought the same thing at first, but the page by default only shows
>> > the
>> > failed cases. The other 1284 cases were OK.
>> >
>> > Gary
>> >
>> >
>> > On Tue, Feb 11, 2014 at 6:07 PM, Clark Boylan <clark.boylan at gmail.com>
>> > wrote:
>> >>
>> >> On Tue, Feb 11, 2014 at 5:52 PM, Gary Duan <garyduan at gmail.com> wrote:
>> >> > Hi,
>> >> >
>> >> > The patch I submitted for L3 service framework integration fails on
>> >> > jenkins
>> >> > test, py26 and py27. The console only gives following error message,
>> >> >
>> >> > 2014-02-12 00:45:01.710 | FAIL: process-returncode
>> >> > 2014-02-12 00:45:01.711 | tags: worker-1
>> >> >
>> >> > and at the end,
>> >> >
>> >> > 2014-02-12 00:45:01.916 | ERROR: InvocationError:
>> >> > '/home/jenkins/workspace/gate-neutron-python27/.tox/py27/bin/python
>> >> > -m
>> >> > neutron.openstack.common.lockutils python setup.py testr --slowest
>> >> > --testr-args='
>> >> > 2014-02-12 00:45:01.917 | ___________________________________ summary
>> >> > ____________________________________
>> >> > 2014-02-12 00:45:01.918 | ERROR:   py27: commands failed
>> >> >
>> >> > I wonder what might be the reason for the failure and how to debug
>> >> > this
>> >> > problem?
>> >> >
>> >> > The patch is at, https://review.openstack.org/#/c/59242/
>> >> >
>> >> > The console output is,
>> >> >
>> >> >
>> >> > http://logs.openstack.org/42/59242/7/check/gate-neutron-python27/e395b06/console.html
>> >> >
>> >> > Thanks,
>> >> >
>> >> > Gary
>> >> >
>> >> >
>> >> > _______________________________________________
>> >> > OpenStack-dev mailing list
>> >> > OpenStack-dev at lists.openstack.org
>> >> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>> >> >
>> >>
>> >> I haven't dug into this too far but
>> >>
>> >>
>> >> http://logs.openstack.org/42/59242/7/check/gate-neutron-python27/e395b06/testr_results.html.gz
>> >> seems to offer some clues. Not sure why the console output doesn't
>> >> show the additional non exit code errors (possibly a nonstandard
>> >> formatter? or a bug?).
>> >>
>> >> Also, cases like this tend to be the test framework completely dying
>> >> due to a sys.exit somewhere or similar. This kills the tests and runs
>> >> only a small subset of them which seems to be the case here.
>> >>
>> >> Clark
>> >>
>> >> _______________________________________________
>> >> OpenStack-dev mailing list
>> >> OpenStack-dev at lists.openstack.org
>> >> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>> >
>> >
>> >
>> > _______________________________________________
>> > OpenStack-dev mailing list
>> > OpenStack-dev at lists.openstack.org
>> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>> >
>>
>> I picked a different neutron master change and it ran 10k py27
>> unittests. Pretty sure the test framework is bailing out early here.
>>
>> Clark
>>
>> _______________________________________________
>> OpenStack-dev mailing list
>> OpenStack-dev at lists.openstack.org
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>



-- 
Robert Collins <rbtcollins at hp.com>
Distinguished Technologist
HP Converged Cloud