[18:41]  * barry waves
[18:46] <smoser> hey barry . 
[18:46] <barry> smoser: hi
[18:46] <smoser> i was lloking at your continuaiton of harlowja's 2to3
[18:46] <harlowja> barry thx for the work btw :)
[18:47] <smoser> i agree that getting tox going is essential.
[18:47] <barry> yw!  i am weird in that i find this kind of thing fun :)
[18:47] <smoser> why did you use contextlib2 ?
[18:48] <barry> smoser: for ExitStack(), which is available in py3 in the stdlib
[18:48] <barry> i might be able to get rid of it if its a problem.  it really helps with managing complicate context managers
[18:49] <barry> sometimes you can avoid it with strategically written addCleanups but it's trickier
[18:49] <harlowja> is that just for testing?
[18:49] <barry> yep
[18:50] <harlowja> would https://github.com/openstack/oslo.db/blob/master/oslo_db/tests/utils.py#L24 suffice?
[18:50] <barry> mostly to properly handle mock patching in the setUps
[18:50] <harlowja> k
[18:50] <barry> (or more technically, unpatching in the tear downs)
[18:50] <harlowja> thinking if its just for testing we can just create something for py2
[18:51] <harlowja> but if its just for testing, then it seems ok
[18:51] <harlowja> less runtime deps :-P
[18:51] <barry> yeah
[18:51] <barry> btw, i think there may be problems with using super() in py2.6
[18:52] <harlowja> how so?
[18:52] <barry> super() only really works with new style classes but unittest.TestCase is an old style class in 2.6
[18:52] <harlowja> ah
[18:52] <harlowja> stupid old-style
[18:52] <barry> stupid 2.6 :)
[18:52] <harlowja> ya
[18:52] <harlowja> lol
[18:53] <smoser> barry: how would you normally get py26 on ubuntu-recent ?
[18:53]  * barry was so happy when he killed that off
[18:53] <barry> smoser: you don't
[18:53] <barry> :)
[18:53] <smoser> but you put 'py26' in the tox.ini file
[18:53] <harlowja> barry ya, sadly some people run rhel6.x (yahoo right now still, moving to rhel7 sometime this year maybe, lol)
[18:53] <barry> well, you can build it from source, but system virtualenv (which tox uses) doesn't work with it
[18:54] <barry> harlowja: i can't tell you what a bane of my existence that's been ;)
[18:54] <harlowja> :)
[18:54] <harlowja> rhel6 or python2.6?
[18:54] <harlowja> or both?
[18:54] <harlowja> lol
[18:54] <barry> smoser: right, my plan was to get it working for 2.6 and 3.4, then to bring up a vm to test 2.6
[18:54] <barry> harlowja: yes, both :-D
[18:54] <smoser> 2.7. yeah.
[18:55] <barry> harlowja, smoser i looked carefully at test_cs_util.py and afaict, even the existing tests never really test anything other than CepkoMock.  they don't touch Cepko afaict
[18:56] <harlowja> ah, cepko, no idea about that, runs away, lol
[18:56] <barry> i basically skip the whole test class in my branch because of that
[18:56] <barry> probably the whole test_cs_util.py file should just be deleted or completely rewritten
[18:57] <barry> i can help y'all integrate coverage at some point if you want :)
[18:57] <harlowja> coverage u say, lol
[18:57] <barry> yeah, i know, crazy talk!
[18:57] <harlowja> thats like how much your body is covered with clothes right
[18:57] <harlowja> ha
[18:57] <barry> harlowja: you know i work from home, right?
[18:57] <harlowja> errrr
[18:57] <harlowja> lol
[18:57] <harlowja> so does smoser :-P
[18:58] <barry> what is this "body covered with clothes" thing of which you speak?
[18:58] <harlowja> true dat
[18:58] <harlowja> lol
[18:58] <barry> :)
[18:58] <smoser> harlowja: is that why you were 'harwlow_at_home' earlier today ?
[18:58] <harlowja> lol
[18:58] <harlowja> quite possibly
[18:59] <barry> i'm fairly sure i now understand how to translate mocker to mock, but there might be corner cases.  so my current task is to convert from mocker to unittest.mock and see if i can get the tests to pass
[18:59] <harlowja> barry so future coverage and stuff, i'm not sure; depends on how we want to focus on https://etherpad.openstack.org/cloud-init-next and https://github.com/cloud-init/cloud-init (and the next hotness)
[18:59] <harlowja> depends on how much we want to put 0.7.x on life-support
[19:00] <harlowja> that decision is above my pay grade, ha
[19:00] <barry> harlowja: yep. understood.  what i'd really suggest is to get the infrastructure in place to do tox+coverage from the start with any rewrite.  don't worry about getting 100% until you have some time to spend on it
[19:00] <smoser> barry: so for the rest of my day, how can i help you on that ?
[19:00] <barry> but knowing is better than not knowing
[19:01] <barry> smoser: help with?  coverage?  the py3 port?
[19:01] <smoser> barry: the tox is in place on cloud-init 2.0, and will make sure to add coverage.
[19:01]  * harlowja not helping u guys get your clothes/coverage on
[19:01] <smoser> from what you've seen on your stuff, where woudl you like me to jump in.
[19:02] <barry> smoser: take a look at the system-image branch.  it's py3-only but should be largely translateable.  and i can help with any details.  http://bazaar.launchpad.net/~ubuntu-system-image/ubuntu-system-image/client/files
[19:02] <smoser> while you admitted to finding 2to3 porting fun, i find it quite annoying.
[19:02] <smoser> ?
[19:02] <barry> smoser: it uses some wacky tox stuff because of the different testing environment requirements between the phone and snappy, but the basics are pretty easy
[19:03] <barry> smoser: yeah.  if you look at the tox.ini and coverage.ini files in that branch, i think you can work out how to integrate coverage
[19:04] <barry> smoser: oh, another thing to consider is switching to nose2 which is much better than nose and the future (nose being essentially eol'd for nose2)
[19:04] <barry> smoser: in that case, look at unittest.cfg from that branch also
[19:04] <barry> and http://bazaar.launchpad.net/~ubuntu-system-image/ubuntu-system-image/client/view/head:/systemimage/testing/nose.py
[19:05] <barry> smoser: steal anything from there you want.  i cargo cult a lot of that stuff between all my projects.
[19:06] <barry> smoser: a simpler, but similar example: https://www.gitorious.org/python-world/python-world/source/a4d9171d8e4a7815d75fd7f8f75bb38418b6c63c:
[19:08] <barry> smoser, harlowja i'll ping y'all when i'm happy with the port.  any and all feedback welcome!  i only want to kill off py2 from the various images.
[19:08] <harlowja> lol
[19:08] <harlowja> cools
[19:10] <smoser> barry: well, how can i help you?
[19:10] <smoser> (thanks fo rhte suggestion of nose2)
[19:11] <barry> smoser: right now, i think i'm okay.  if i get stuck or have questions, i'l ping you.  i want to have a branch ready for review by eow (maybe eod if i get lucky).
[19:11] <barry> smoser: thanks!
[19:11] <smoser> ok then. i'll get out of yoru way :)
[19:12] <smoser> the other no-python-3 thing is oauth
[19:12] <smoser> cheetah really is not needed.
[19:12] <smoser> the tests my insist on it, but in running practice its only used if user needs it (which is unlikely)
[19:14] <barry> smoser: oh, that would actually be helpful.  you should switch to oauthlib.  the api is different enough that it takes some thought, but oauthlib is py3 compatible and actually maintained for the last 6 years :)
[19:14] <barry> (oauth being abandonware since 2009)
[19:14] <smoser> barry: yes. i know about oathlib. 
[19:14] <smoser> yeah. 
[19:14] <barry> i saw the oauth imports, but promptly erased them :)
[19:14] <barry> from my brain
[19:15] <barry> i've done that port before, it's not too difficult but it's not entirely trivial
[19:15] <harlowja> only for maas i think right?
[19:16] <barry> harlowja: i can't answer that, since it would require me to actually understand the code :)
[19:16] <smoser> right
[19:16]  * harlowja pretty sure thats right, ha
[19:16] <harlowja> why u no understand all the code yet, lol
[19:16] <harlowja> ha
[19:17]  * barry goes clicky clacky
[19:17] <harlowja> fasterrrrr!
[20:02] <harmw> harlowja: what do you guys use for storage in openstack deploymnts?
[20:02] <harmw> (or you smoser)
[20:02] <harlowja> harmw ah; so this is tricky area for yahoo
[20:03] <harlowja> yahoo has a bunch of cloudy storage stuff that already exists and so block-storage isn't the top contendor for us
[20:03] <harlowja> http://en.wikipedia.org/wiki/Yahoo_Sherpa is one such system
[20:03] <harlowja> *closed-source
[20:03] <harmw> oh ok
[20:03] <harlowja> ceph is getting more adoption but most architects i think still prefer using the other stuff thats built if possible (which already works in mostly cloudy manner)
[20:04] <harmw> other stuff like.. :)
[20:04] <harlowja> hmmm, there are some other db-aas stuff that people built; so i think thats another one
[20:04] <harmw> i want a 2 perhaps 3 node setup, just not settled on the storage part
[20:04] <harlowja> block-storage though is more of a slow-up-and-coming thing
[20:04] <harlowja> *since this existing stuff already exists
[20:04] <harlowja> http://en.wikipedia.org/wiki/MObStor is another one harmw 
[20:05] <harlowja> probably some other ones that are not really documented online (but are mentioned somewhere/somehow by leaks, lol)
[20:05] <harmw> hehe
[20:05] <harlowja> sherpa though i think is the thing that most people are getting directed to
[20:06] <harlowja> *or that db-aas stuff; since alot of groups still need sql-like-stuff
[20:06] <harmw> perhaps I'm just gonna use LVM here
[20:06] <harlowja> http://en.wikipedia.org/wiki/Yahoo!_Query_Language is also recommended (its a query-as-a-service  in a way)
[20:07] <harlowja> so anyways, different things :-P
[20:07] <harmw> yup
[20:07] <harmw> useless answers, again :P
[20:07] <harmw> damn
[20:07] <harlowja> :)
[20:08] <harmw> im not even sure if I should do TripleO 
[20:08] <harmw> it looks very cool
[20:08] <harlowja> for our current vms (due to above mentioned exsting services) its mostly local-disk (and treat your vms as cattle, put needed data in sherpa or elsewhere...)
[20:08] <harlowja> we'll get to ceph eventually and make this more accessible (but have to convince people that is needed over things like sherpa...)
[20:09] <harlowja> ^ is part of the battle/pain, lol
[20:09] <harmw> :)