| *** tetsuro has joined #openstack-placement | 00:11 | |
| *** sean-k-mooney has quit IRC | 00:44 | |
| *** sean-k-mooney has joined #openstack-placement | 00:58 | |
| openstackgerrit | Merged openstack/placement master: Remove keystoneauth1 opts from placement config group https://review.openstack.org/619299 | 02:56 |
|---|---|---|
| openstackgerrit | Chris Dent proposed openstack/placement master: Remove [keystone] config options from placement https://review.openstack.org/620412 | 03:08 |
| openstackgerrit | Chris Dent proposed openstack/placement master: WIP: Stop using global oslo_config https://review.openstack.org/619121 | 03:08 |
| *** cdent has quit IRC | 03:28 | |
| *** diga has joined #openstack-placement | 03:47 | |
| *** diga has quit IRC | 04:49 | |
| *** e0ne has joined #openstack-placement | 05:37 | |
| *** e0ne has quit IRC | 05:42 | |
| *** e0ne has joined #openstack-placement | 05:44 | |
| *** e0ne has quit IRC | 05:45 | |
| *** alex_xu has quit IRC | 07:50 | |
| *** e0ne has joined #openstack-placement | 07:54 | |
| *** alex_xu has joined #openstack-placement | 07:55 | |
| openstackgerrit | Tetsuro Nakamura proposed openstack/placement master: Add alembic version stamp capability to the DB https://review.openstack.org/620216 | 08:01 |
| openstackgerrit | Tetsuro Nakamura proposed openstack/placement master: Set root_provider_id in the database https://review.openstack.org/619126 | 08:04 |
| *** e0ne has quit IRC | 08:37 | |
| *** tssurya has joined #openstack-placement | 08:41 | |
| *** tetsuro has quit IRC | 09:03 | |
| openstackgerrit | Debo Zhang proposed openstack/placement master: Fix typo https://review.openstack.org/620541 | 09:21 |
| openstackgerrit | Ghanshyam Mann proposed openstack/nova-specs master: Spec for API inconsistency cleanup https://review.openstack.org/603969 | 09:22 |
| openstackgerrit | Ghanshyam Mann proposed openstack/nova-specs master: Spec for API inconsistency cleanup https://review.openstack.org/603969 | 09:23 |
| *** ttsiouts has joined #openstack-placement | 09:59 | |
| *** ttsiouts has quit IRC | 10:02 | |
| *** ttsiouts has joined #openstack-placement | 10:03 | |
| *** ttsiouts has quit IRC | 10:03 | |
| *** ttsiouts has joined #openstack-placement | 10:03 | |
| *** ttsiouts has quit IRC | 10:06 | |
| *** ttsiouts has joined #openstack-placement | 10:07 | |
| *** cdent has joined #openstack-placement | 10:11 | |
| *** ttsiouts has quit IRC | 10:12 | |
| cdent | morning gibi | 10:12 |
| *** ttsiouts has joined #openstack-placement | 10:13 | |
| gibi | cdent: morning | 10:13 |
| *** ttsiouts has quit IRC | 10:14 | |
| *** e0ne has joined #openstack-placement | 10:14 | |
| *** ttsiouts has joined #openstack-placement | 10:15 | |
| *** ttsiouts has quit IRC | 10:18 | |
| *** ttsiouts has joined #openstack-placement | 10:19 | |
| *** ttsiouts has quit IRC | 10:21 | |
| *** ttsiouts has joined #openstack-placement | 10:22 | |
| *** ttsiouts has quit IRC | 10:22 | |
| openstackgerrit | Merged openstack/placement master: Remove [keystone] config options from placement https://review.openstack.org/620412 | 10:44 |
| cdent | gibi: I fixed the two easy to fix test failures on the nova using external placement change but I'm basically stumped at this point. I'm running the placement tests in a loop now, just to be sure they don't have anything lingering, but I don't think that's a factor. | 10:46 |
| cdent | So I'm now really sure how to proceed from here, other than the eventlet poking you're planning to do. Any other ideas? | 10:47 |
| cdent | I'll get to the alembic and lock fixes on the placement side in a while | 10:48 |
| cdent | For giggles i ran the placement tests --uniti-failure. I got to >100000 tests and gave up | 11:06 |
| *** ttsiouts has joined #openstack-placement | 11:13 | |
| gibi | cdent: I lost the morning runs due to a mistake. I thought I run the test with a lock in @run_once but I added -r to the tox command that re-coloned placement in the env :/ | 11:16 |
| gibi | in theory run_once in in an eventlet environment allows calling the wrapped func more than once, if during the first call an eventlet switch happens | 11:17 |
| gibi | as run_once set the wrapped.called = True after the func returned | 11:17 |
| gibi | (or raised) | 11:18 |
| *** ttsiouts has quit IRC | 11:18 | |
| openstackgerrit | Brin Zhang proposed openstack/nova-specs master: Support for changing deleted_on_termination after boot https://review.openstack.org/580336 | 11:33 |
| gibi | cdent: I've also tried to record who calls the configure method but as soon as I add some extra code to that place the problem disappears so I was not able to catch the first caller when there are more than one calls :/ Most probalby adding more code there changes the timing of the race | 11:37 |
| cdent | gibi: ugh. This is _so_ much fun... | 11:41 |
| gibi | cdent: we will learn from it... if we figure out what is wrong | 11:42 |
| cdent | I wish we could turn eventlet all the way off in the nova tests. It's the source of much pain. | 11:42 |
| cdent | Oh yeah, the learning is good | 11:42 |
| *** e0ne has quit IRC | 12:09 | |
| *** ttsiouts has joined #openstack-placement | 12:11 | |
| *** ttsiouts has quit IRC | 12:12 | |
| *** ttsiouts has joined #openstack-placement | 12:13 | |
| *** ttsiouts has quit IRC | 12:17 | |
| gibi | just to increase the fun factor with PS 25 I have been running the tests without failure in the last 40 minutes | 12:27 |
| * cdent shivers | 12:29 | |
| *** jaypipes has joined #openstack-placement | 12:31 | |
| cdent | moin jaypipes | 12:31 |
| jaypipes | cdent: hey Chris, how are ya? :) | 12:33 |
| cdent | I'm swinging from disappointed and meh about everything to omg, we're fixing it! | 12:34 |
| jaypipes | cdent: heh, I'm not swinging. | 12:35 |
| jaypipes | cdent: just seem to be in a permanent state of disappointment of late. | 12:35 |
| cdent | did you see my removing global config email? | 12:35 |
| cdent | I included you on it beause it might shine a small bit of light | 12:35 |
| cdent | I'm sure I'll swing the other way soon enough, but for now it's a brief moment of nice | 12:36 |
| jaypipes | cdent: heh, just sent a reply. great stuff. will review today. | 12:37 |
| cdent | I've fixed the "need to be fixed bits" and am just now making sure it is all kosher | 12:37 |
| cdent | thanks | 12:38 |
| jaypipes | np, exciting to see it. | 12:38 |
| gibi | cdent: OK, I just needed to be patient. Now failed again | 12:38 |
| *** ttsiouts has joined #openstack-placement | 12:38 | |
| cdent | gibi: gah! | 12:39 |
| cdent | I don't really have any ideas. Do you have a working theory of what's going on? | 12:40 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Stop using global oslo_config https://review.openstack.org/619121 | 12:41 |
| cdent | that's ^ passing everything locally now, including the opportunistic tests, we'll see what tempest thinks of it | 12:41 |
| gibi | cdent: this was still without to lock in run_once, so now I'm adding the lock and running again | 12:42 |
| cdent | ah, okay | 12:42 |
| cdent | that's a relief | 12:42 |
| cdent | I thought you were saying that the lock didn't help | 12:42 |
| gibi | cdent: now I have a baseline that ps 25 still fails, and needs at least and hour of continuous run to reproduec | 12:42 |
| cdent | how many total tests does an hour produce on your beastly machine? | 12:43 |
| gibi | 24 full functional run | 12:45 |
| jaypipes | cdent: I'm unfortunately unfamiliar with the tox siblings stuff. Does adding a required-projects in the .zuul.yaml essentially require setting tox_install_siblings? Also, what specifically does tox_install_siblings *do*? :) Looking at the YAML for the ansible task here doesn't really indicate anything: https://github.com/openstack-infra/zuul-jobs/blob/master/roles/tox/tasks/siblings.yaml. Plus, googling for "tox siblings" gives me results | 12:53 |
| jaypipes | for "toxic siblings" which is funny, but non-useful. | 12:53 |
| cdent | jaypipes: ah yeah, it is all a bit obscure. basically it goes like this: | 12:54 |
| cdent | required projects lists the things that zuul is going to git clone early in its process | 12:54 |
| *** ttsiouts has quit IRC | 12:56 | |
| cdent | if tox siblings is turned on, then at some stage in the process zuul will look at the packages that are in the virtualenv for the jobs it is running and if any of them are provided by and of the things is had cloned early, it will re-install those packages into the virtenv, from the local git repo | 12:56 |
| *** ttsiouts has joined #openstack-placement | 12:56 | |
| cdent | this means that you will either get master or whateven the depends-on version of that thing (in this case placement) in the test environment | 12:56 |
| jaypipes | cdent: got it. thanks for the excellent explanation! | 13:03 |
| *** e0ne has joined #openstack-placement | 13:05 | |
| jaypipes | cdent: +269, -17988 <-- nice. | 13:07 |
| *** ttsiouts has quit IRC | 13:12 | |
| *** ttsiouts has joined #openstack-placement | 13:13 | |
| *** ttsiouts has quit IRC | 13:18 | |
| *** ttsiouts has joined #openstack-placement | 13:33 | |
| jaypipes | cdent: sorry, getting back to a review on the nova patch for removing placement stuff... could you explain to me what the from __future__ import absolute_import accomplishes in https://review.openstack.org/#/c/617941/25/nova/tests/functional/api_paste_fixture.py? | 13:35 |
| jaypipes | cdent: is there some placement-y module that is shadowing one of the imports there? | 13:36 |
| cdent | jaypipes: catching up, was eating | 13:40 |
| * cdent reads | 13:40 | |
| jaypipes | cdent: ah, maybe the fixtures... | 13:41 |
| cdent | yeah, it's the fixtures | 13:41 |
| cdent | I moved some fixtures out of nova/test/fixtures into that dir | 13:41 |
| jaypipes | cdent: hmm, ok. I thought absolute_import was the default >= py27? | 13:42 |
| cdent | well, if you don't have that line, in 27, it will try to load the wrong thing, so I guess not? | 13:42 |
| *** diga has joined #openstack-placement | 13:42 | |
| jaypipes | cdent: :) | 13:42 |
| cdent | I recall reading something about it, but the details escape me | 13:43 |
| cdent | jaypipes: https://bugs.python.org/issue14494 | 13:49 |
| jaypipes | cdent: on https://review.openstack.org/#/c/617941/25/nova/tests/functional/regressions/test_bug_1679750.py, is that what you want long-term, or is that more of a temporary solution? My guess is you're trying to get full separation of placement modules right? so, not import'ing anything from placement directly into nova tests, yeah? | 13:50 |
| jaypipes | cdent: ah, cool, good hunting on the absolute import thing | 13:50 |
| cdent | jaypipes: that probalby ought to be a short term thing, but that test is "special" and I couldn't think of a good way to do it more cleanly | 13:51 |
| jaypipes | cdent: yeah, figured as much, just wanted to check with ya | 13:51 |
| gibi | cdent: I tried the locking and I still fail. I now think that run_once has some hidden bug itself that we are overlooking. I will try to remove run_once and implement a minimalistic guard around content of configure() directly | 13:52 |
| cdent | gibi: :( Thanks for pluggin away. | 13:53 |
| cdent | good that you have some meaty hardware. Are there any avenues I should explore? | 13:53 |
| gibi | cdent: this is now so strange that I cannot stop. fortunately I have still an idea (the removeal of the complex run_once) to try | 13:53 |
| cdent | jaypipes: if/when it needs a respin, I can stick a TODO in there some where | 13:54 |
| cdent | "now so strange that I cannot stop" is my new tshirt | 13:54 |
| gibi | cdent: if you can put a fresh eye the run_once decorator and the reset() function in the fixtures that calls the configure.reset() maybe you can spot something. Besides that I don't know | 13:55 |
| *** diga has quit IRC | 13:56 | |
| cdent | k, I'll try to do some thinking, but after last night I'm not sure I can promise much today | 13:56 |
| gibi | cdent: no worries most probablty we have already looked too much into this. | 13:57 |
| cdent | in related news, I was running nova master in an --until-fail and it just did | 13:58 |
| cdent | I'll see if that lends a clue | 13:58 |
| jaypipes | cdent: not a prob | 14:03 |
| cdent | sean-k-mooney: you around? you might have some ideas on the test failures gibi and I are seeing when using external placement with nova functional tests | 14:07 |
| *** ttsiouts has quit IRC | 14:08 | |
| *** ttsiouts has joined #openstack-placement | 14:09 | |
| *** ttsiouts_ has joined #openstack-placement | 14:11 | |
| sean-k-mooney | i am | 14:12 |
| sean-k-mooney | im not sure if i will but i can take a look if you send me a link | 14:12 |
| jaypipes | cdent: https://review.openstack.org/#/c/617941/ reviewed. good stuff. not +1 b/c of a question, but close. | 14:13 |
| cdent | it's this patch, sean-k-mooney , which if you run forever will eventually have errors as indicated in the pastes that gibi has inked from it: https://review.openstack.org/#/c/617941/ | 14:13 |
| *** ttsiouts has quit IRC | 14:13 | |
| cdent | thanks jaypipes, looking | 14:13 |
| cdent | jaypipes: there's commentary on that topic between mriedem and on ps 21 and at http://eavesdrop.openstack.org/irclogs/%23openstack-nova/%23openstack-nova.2018-11-26.log.html#t2018-11-26T15:44:32 | 14:15 |
| sean-k-mooney | cdent: gibi so calling reset currently does nothing except clear the ran flag as there is no cleanup handeler to actully reset the garded transaction factory | 14:15 |
| cdent | but basically: yeah, we need to figure out that out, jaypipes | 14:15 |
| cdent | sean-k-mooney: there is in the Database fixture placement/tests/fixtures.py | 14:16 |
| sean-k-mooney | cdent: are you pasing in write contexts via decoratrors anywhere | 14:16 |
| * sean-k-mooney looks | 14:16 | |
| cdent | yes, we covered that before | 14:17 |
| cdent | when making run once, there all over the place, idiomatic | 14:17 |
| sean-k-mooney | then reset wont work | 14:17 |
| sean-k-mooney | the decorator will never be rebound | 14:17 |
| cdent | sean-k-mooney: have a look at the fixture, it works fine on the placement side | 14:17 |
| cdent | where it seems to go wrong is when eventlet gets involved once nova is also involved | 14:18 |
| *** guilhermesp has joined #openstack-placement | 14:19 | |
| sean-k-mooney | https://github.com/openstack/placement/blob/master/placement/tests/fixtures.py#L43 even with dispose i think that might be undefied behavior | 14:20 |
| cdent | when that code was added zzzeek was in the mix. It was the only thing that worked in a situation where we wanted three different potential engines-types in the same process | 14:21 |
| sean-k-mooney | but in any case you are saying this works fine in placement but then when the same fixture is used in nova its failing | 14:22 |
| cdent | it's racing. rarely. | 14:22 |
| * cdent finds a recent paste | 14:22 | |
| sean-k-mooney | well that kind of make sense | 14:23 |
| cdent | line 102 on http://paste.openstack.org/show/736292/ | 14:23 |
| *** mriedem has joined #openstack-placement | 14:23 | |
| *** ttsiouts_ has quit IRC | 14:23 | |
| sean-k-mooney | this should have a lock around it i think https://github.com/openstack/placement/blob/master/placement/util.py#L426 | 14:23 |
| cdent | I think gibi tried some form of locking, but I'm not sure if was right there. gibi? | 14:23 |
| *** ttsiouts has joined #openstack-placement | 14:24 | |
| sean-k-mooney | we proably need to lock around https://github.com/openstack/placement/blob/master/placement/util.py#L426-L435 and https://github.com/openstack/placement/blob/master/placement/util.py#L448-L452 | 14:25 |
| cdent | on the placement side it's not an issue because we never go co-operative. If we were in the mood for another huge change, getting evently out of the nova tests would be handy, but I haven't got the stamina | 14:25 |
| sean-k-mooney | this was writtend with the expection that this was single threaded but eventlest could case a yeild in the invocation of func | 14:26 |
| sean-k-mooney | let me try a small patch with a lock in those two places and see if we can fix the race | 14:27 |
| cdent | cool, thanks | 14:27 |
| cdent | ping me when you've got something and I can try testing it | 14:28 |
| sean-k-mooney | cdent: is there a simple way for me to test it locally by the way or just run the functional tests repetatedly and see if they fail | 14:28 |
| cdent | if you could base it on https://review.openstack.org/#/c/619121/ that's probably best | 14:28 |
| *** ttsiouts has quit IRC | 14:28 | |
| cdent | i learned the following from gibi: tox -e functional -r -- --random --until-failure | 14:29 |
| sean-k-mooney | does random jsut run tests in a random order? | 14:29 |
| sean-k-mooney | cdent: i can base it on that on the placement side but should i base it on https://review.openstack.org/#/c/617941/ on the nova side? | 14:31 |
| cdent | yes | 14:31 |
| cdent | you shouldn't need any changes on the nova side, just the placement side, but in order to use the placement side you need that nova patch | 14:31 |
| sean-k-mooney | oh your right in that case ill checkout https://review.openstack.org/#/c/617941/ and use a local copy of placement | 14:32 |
| * gibi sitting on a meeting | 14:33 | |
| gibi | cdent, sean-k-mooney: I tried locking this way http://paste.openstack.org/show/736318/ | 14:36 |
| gibi | cdent, sean-k-mooney: and also tried and extra lock around fixtures.reset() | 14:37 |
| sean-k-mooney | .. that was basically what i was going to do. | 14:37 |
| gibi | cdent, sean-k-mooney: and I was still able to reproduce the problem | 14:37 |
| *** ttsiouts has joined #openstack-placement | 14:39 | |
| sean-k-mooney | gibi: that ptache should have prevent any potential race | 14:40 |
| cdent | I wonder if our understanding of the problem is way off and we are following a red herring. I haven't got any alternate ideas though. | 14:40 |
| gibi | sean-k-mooney: yeah, that was my thought as well. I'm now trying to replace the complex run_once with a simple if (and a lock) in the configure() call as I feel the run_one is more complex than I'm happy with | 14:41 |
| sean-k-mooney | i dont think that will help | 14:41 |
| gibi | sean-k-mooney: I have no better idea right now so I try this | 14:41 |
| sean-k-mooney | i dont think this is an issue with the run_once decorator * onece the lock is added | 14:41 |
| sean-k-mooney | sure | 14:42 |
| sean-k-mooney | i do have once question | 14:42 |
| sean-k-mooney | do we guarentee that withing hte test runner that we will not interleave two tests in the same process | 14:42 |
| cdent | sean-k-mooney: with eventlet in place all bets are off, no? | 14:43 |
| gibi | sean-k-mooney: I tend to assume that one test per process runs at once | 14:43 |
| sean-k-mooney | cdent: that what i was getting too with eventlet im not sure gibis assumetion woudl hold | 14:44 |
| gibi | could there be a not joined eventlet lingering around from a previous test while the next tests start runnig? | 14:44 |
| sean-k-mooney | i dont know. but coudl we move reset to the setUP function | 14:48 |
| sean-k-mooney | so call it here | 14:48 |
| sean-k-mooney | https://github.com/openstack/placement/blob/master/placement/tests/fixtures.py#L68 | 14:48 |
| cdent | sean-k-mooney: probably worth trying | 14:50 |
| openstackgerrit | Balazs Gibizer proposed openstack/placement master: DNM: replace run_once with inline code https://review.openstack.org/620617 | 14:51 |
| cdent | sean-k-mooney: presumably to be most tidy we'd need reset both at the setUp and cleanUp | 14:52 |
| sean-k-mooney | i guess. doing it isn setup is say ill start form a know good place. doing it is cleanup is i made a mess let me fix that for you | 14:53 |
| cdent | right, but we can't predict what nova will do so returning to a default state at the end is probably best | 14:54 |
| sean-k-mooney | where are you useing the placement fixture in nova by the way | 14:55 |
| sean-k-mooney | here https://review.openstack.org/#/c/617941/25/nova/tests/functional/fixtures.py? | 14:55 |
| cdent | all over the place, but notably with any test that tries to schedule an instance (thus many places) | 14:56 |
| sean-k-mooney | what i ment to ask is how are teh setup and cleanup functions called | 14:56 |
| cdent | the fixture is used in the usual self.useFixture(...) way | 14:57 |
| sean-k-mooney | ah they are called like this https://review.openstack.org/#/c/617941/25/nova/tests/functional/regressions/test_bug_1550919.py@149 | 14:57 |
| cdent | my conection to gerrit is interrupted somewhere in the middle | 14:58 |
| sean-k-mooney | oh well it was jsut self.useFixture(func_fixtures.PlacementFixture()) in the tests setUp function | 14:59 |
| gibi | sean-k-mooney, cdent: fyi, I've started running the functional test in loop with https://review.openstack.org/620617 I expect to run in 2 x 24 full functional runs at least | 15:06 |
| cdent | i'm trying a --until-failure on reset() in both setup and cleanup | 15:06 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Stop using global oslo_config https://review.openstack.org/619121 | 15:14 |
| cdent | meh, reset() on both sides does not fix it, but still probably a useful thing to have | 15:26 |
| gibi | cdent: my trial still runs (now running the 11th of full functional run) | 15:33 |
| cdent | 15:34 | |
| * cdent crosses fingers | 15:34 | |
| openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: Support volume-backed server rebuild https://review.openstack.org/532407 | 15:59 |
| cdent | mriedem: are you aware of any devstack merge fallout? I haven't heard any. | 16:09 |
| mriedem | not yet | 16:10 |
| gibi | cdent: the test run passed the 30th run without failure | 16:15 |
| cdent | gibi: a) yay! b) my reset experiment above was likely insufficient, going to try a slight modification | 16:15 |
| *** ttsiouts has quit IRC | 16:17 | |
| *** ttsiouts has joined #openstack-placement | 16:18 | |
| cdent | on b) the config is in init not setUp, that's wrong | 16:18 |
| * cdent tries again | 16:18 | |
| *** ttsiouts has quit IRC | 16:22 | |
| *** ttsiouts has joined #openstack-placement | 16:28 | |
| *** ttsiouts has quit IRC | 16:39 | |
| *** ttsiouts has joined #openstack-placement | 16:40 | |
| *** ttsiouts has quit IRC | 16:41 | |
| *** tssurya has quit IRC | 16:45 | |
| cdent | "wrong" even if it doesn't fix the problem | 16:46 |
| cdent | gibi: your change on https://review.openstack.org/#/c/620617/ isn't going to cause a failure, it will only warn, or move the failure to somewhere else, correct? Which might be fine. | 16:55 |
| gibi | cdent: it logs the same warning as the run_once decorator logged if the configure was called more than once | 16:56 |
| gibi | cdent: my intention was to simply move the behaviour from run_once directly into configure() | 16:57 |
| cdent | the thing in the previous fails is not a warning, it's an exception raised by oslo_db: http://paste.openstack.org/show/736292/ (line 102) | 16:57 |
| cdent | because configure was called | 16:57 |
| cdent | what your change does is make it so configure won't be called, and then logs a warning | 16:57 |
| gibi | cdent: what I really changed is that I'm not depending on yet another global(ish) flag 'called' but depending on the internal of sqlaclhemy to know if configure was called before | 16:58 |
| cdent | yes, that's probably a nice thing, but we haven't got the same failure surface now, so we don't know if we've "fixed it", because "it" is now different | 16:58 |
| gibi | cdent: in my mind when run_once let the configure to be called the second time was a bug | 16:58 |
| cdent | right, but the way the placement side is set up, if configure is attempted to be called a second time, that's a mistake | 16:59 |
| cdent | which we should know about | 16:59 |
| cdent | we don't want to simply say "oh you tried, that's okay" | 17:00 |
| gibi | cdent: I can make the else: branch raise in configure | 17:00 |
| cdent | it's probably okay in the context of a running service, where the error showed up before, but in tests we want to know | 17:00 |
| cdent | so it's confusing... | 17:00 |
| gibi | btw the test just passed the 48th time without problem | 17:01 |
| cdent | yeah, my point is: the problem can't happen with your code | 17:01 |
| *** e0ne has quit IRC | 17:02 | |
| cdent | whether that means we are masking something else is unclear | 17:02 |
| gibi | OK, so during testing we want to fail if somebody called configure twice | 17:03 |
| gibi | without reset() | 17:03 |
| cdent | unless we've explicit done the reset | 17:03 |
| cdent | yeah | 17:04 |
| gibi | yeah | 17:04 |
| cdent | :) | 17:04 |
| gibi | OK, I can tune the code tomorrow to fail on the else branch | 17:04 |
| cdent | it's appearing like the change I have in progress is going to change the timing of things as well, making the race less likely, and is a right fix, so I'll push that up too. | 17:05 |
| cdent | we're going to end up with lots of fixed and better stuff and never be quite sure which thing(s) fixed it | 17:05 |
| openstackgerrit | Balazs Gibizer proposed openstack/placement master: DNM: replace run_once with inline code https://review.openstack.org/620617 | 17:05 |
| cdent | but better is good | 17:05 |
| gibi | I will run this during the night ^^ | 17:06 |
| cdent | cool, thanks | 17:07 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Stop using global oslo_config https://review.openstack.org/619121 | 17:10 |
| openstackgerrit | Chris Dent proposed openstack/placement master: DNM: replace run_once with inline code https://review.openstack.org/620617 | 17:11 |
| openstackgerrit | Merged openstack/placement master: Fix typo https://review.openstack.org/620541 | 17:11 |
| gibi | cdent: already the first run failed with the ValueError I added | 17:14 |
| gibi | cdent: but now in 20 places | 17:15 |
| cdent | data! | 17:15 |
| cdent | gibi: try rebasing on my updated change, which fixes the Database fixture to be more correct | 17:15 |
| gibi | cdent: this is the failure now http://paste.openstack.org/show/736328/ | 17:16 |
| gibi | cdent: do you mean using the ps 3 from my change that you rebased | 17:16 |
| gibi | ? | 17:16 |
| sean-k-mooney | what would it take to kill the global and just pass the context manger down to things that need it | 17:16 |
| sean-k-mooney | im assuming it not trivial but long term that might be for the best | 17:17 |
| cdent | gibi: yes, it moves when configure is called, to a more correct place | 17:17 |
| cdent | (after reset) | 17:17 |
| gibi | cdent: OK | 17:17 |
| gibi | but now I have to really leave | 17:17 |
| gibi | talk to you tomorrow | 17:17 |
| cdent | sean-k-mooney: I think it would still be needed for the opportunistic tests, so much of the facade stuff expects a sort-of-global engine | 17:18 |
| cdent | and it is a nice convenience and the "only" reason we are having difficulty is because we are trying to do an until now abnormal thing with cross-project functional tests | 17:20 |
| *** e0ne has joined #openstack-placement | 19:07 | |
| *** e0ne has quit IRC | 19:15 | |
| * cdent taps out | 19:17 | |
| *** cdent has quit IRC | 19:17 | |
| *** e0ne has joined #openstack-placement | 19:17 | |
| *** e0ne has quit IRC | 19:18 | |
| openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: Support volume-backed server rebuild https://review.openstack.org/532407 | 19:40 |
| *** e0ne has joined #openstack-placement | 20:19 | |
| *** e0ne has quit IRC | 20:20 | |
| *** openstackgerrit has quit IRC | 20:36 | |
| *** e0ne has joined #openstack-placement | 20:40 | |
| *** e0ne has quit IRC | 20:49 | |
| *** e0ne has joined #openstack-placement | 21:21 | |
| *** e0ne has quit IRC | 21:21 | |
| *** e0ne has joined #openstack-placement | 21:26 | |
| *** e0ne has quit IRC | 21:27 | |
| *** takashin has joined #openstack-placement | 21:50 | |
| *** openstackgerrit has joined #openstack-placement | 21:51 | |
| openstackgerrit | Merged openstack/nova-specs master: Support volume-backed server rebuild https://review.openstack.org/532407 | 21:51 |
| *** N3l1x has quit IRC | 23:17 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!