| *** tetsuro has joined #openstack-placement | 00:16 | |
| *** mriedem has quit IRC | 00:29 | |
| *** bhagyashris has joined #openstack-placement | 00:57 | |
| *** openstackgerrit has joined #openstack-placement | 01:24 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: DNM: Don't use OVO with allocation candidates https://review.openstack.org/636631 | 01:24 |
|---|---|---|
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO for Allocation and AllocationList https://review.openstack.org/636694 | 01:24 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO for Usage and UsageList https://review.openstack.org/636695 | 01:24 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO in Consumer object https://review.openstack.org/636801 | 01:24 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO in Project object https://review.openstack.org/636802 | 01:24 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO in User object https://review.openstack.org/636803 | 01:24 |
| openstackgerrit | Chris Dent proposed openstack/placement master: WIP: Don't use OVO in ResourceClass and ResourceClassList https://review.openstack.org/636804 | 01:24 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO in Trait and TraitList objects https://review.openstack.org/636805 | 01:24 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO for Inventory and InventoryList https://review.openstack.org/636806 | 01:24 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO with ResourceProvider and ResourceProviderList https://review.openstack.org/636807 | 01:24 |
| *** cdent has quit IRC | 01:42 | |
| *** tetsuro has quit IRC | 05:40 | |
| nguyenhai_ | thanks cdent | 05:48 |
| *** tetsuro has joined #openstack-placement | 06:03 | |
| *** tetsuro has quit IRC | 06:05 | |
| nguyenhai_ | after reset the apache2, I met this error: apache2 : Depends: apache2-bin (= 2.4.18-2ubuntu3.9) but 2.4.29-1ubuntu4.2 is to be installed | 06:19 |
| nguyenhai_ | please help | 06:19 |
| *** e0ne has joined #openstack-placement | 06:29 | |
| nguyenhai_ | apache2 : Depends: apache2-bin (= 2.4.18-2ubuntu3.9) but it is not going to be installed | 06:44 |
| *** e0ne has quit IRC | 06:46 | |
| nguyenhai_ | I solved it by modifying the sources.list to compatible with the upgrade Ubuntu :) Thanks all | 07:08 |
| openstackgerrit | Tetsuro Nakamura proposed openstack/placement master: Trivial: return empty set instead of list https://review.openstack.org/636846 | 07:37 |
| *** e0ne has joined #openstack-placement | 08:05 | |
| *** e0ne has quit IRC | 08:08 | |
| *** e0ne has joined #openstack-placement | 08:18 | |
| *** e0ne has quit IRC | 08:18 | |
| *** e0ne has joined #openstack-placement | 08:29 | |
| *** e0ne has quit IRC | 08:35 | |
| *** helenafm has joined #openstack-placement | 08:46 | |
| *** helenafm has quit IRC | 08:50 | |
| *** tssurya has joined #openstack-placement | 08:54 | |
| *** helenafm has joined #openstack-placement | 08:58 | |
| *** bhagyashris has quit IRC | 09:42 | |
| *** tetsuro has joined #openstack-placement | 09:50 | |
| *** tetsuro has quit IRC | 09:51 | |
| *** cdent has joined #openstack-placement | 10:13 | |
| openstackgerrit | sahid proposed openstack/osc-placement master: Change python3.5 job to python3.7 job on Stein+ https://review.openstack.org/610074 | 10:21 |
| openstackgerrit | sahid proposed openstack/os-traits master: Change python3.5 job to python3.7 job on Stein+ https://review.openstack.org/610065 | 10:22 |
| *** helenafm has quit IRC | 10:38 | |
| *** tetsuro has joined #openstack-placement | 10:41 | |
| *** tetsuro has quit IRC | 10:42 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: DNM: Don't use OVO with allocation candidates https://review.openstack.org/636631 | 10:49 |
| *** helenafm has joined #openstack-placement | 11:08 | |
| *** vdrok_ has quit IRC | 11:44 | |
| *** vdrok has joined #openstack-placement | 11:48 | |
| *** tetsuro has joined #openstack-placement | 11:51 | |
| *** tetsuro has quit IRC | 12:16 | |
| *** tetsuro has joined #openstack-placement | 12:19 | |
| *** tetsuro has quit IRC | 12:21 | |
| *** tetsuro has joined #openstack-placement | 12:24 | |
| cdent | sigh, running the placement unit tests takes about 1/100th of the time it takes to do whatever tox and stestr are getting up to | 12:46 |
| *** tetsuro has quit IRC | 12:56 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: DNM: Don't use OVO with allocation candidates https://review.openstack.org/636631 | 12:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO for Allocation and AllocationList https://review.openstack.org/636694 | 12:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO for Usage and UsageList https://review.openstack.org/636695 | 12:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO in Consumer object https://review.openstack.org/636801 | 12:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO in Project object https://review.openstack.org/636802 | 12:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO in User object https://review.openstack.org/636803 | 12:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO in ResourceClass and ResourceClassList https://review.openstack.org/636804 | 12:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO in Trait and TraitList objects https://review.openstack.org/636805 | 12:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO for Inventory and InventoryList https://review.openstack.org/636806 | 12:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Don't use OVO with ResourceProvider and ResourceProviderList https://review.openstack.org/636807 | 12:57 |
| *** mriedem has joined #openstack-placement | 13:39 | |
| *** s10 has joined #openstack-placement | 13:48 | |
| *** cdent has quit IRC | 14:04 | |
| *** cdent has joined #openstack-placement | 14:15 | |
| cdent | jaypipes: in addition to other responses on to your message on the evaluation thread, I pointed to some ovo data (with code) | 14:48 |
| cdent | thanks very much for getting on that thread | 14:48 |
| cdent | it's kind of relief when email actually happens | 14:48 |
| jaypipes | cdent: thx for writing it! :) | 14:48 |
| openstackgerrit | Merged openstack/nova-specs master: Add more detail and fix code block formatting https://review.openstack.org/628702 | 15:14 |
| *** s10 has quit IRC | 15:20 | |
| efried | cdent: I think we can fix your tox perf woes, at least some of them, by porting https://review.openstack.org/636378 and https://review.openstack.org/582392 to the placement repo (and you installing tox>=3.11 locally). Gimme a few, I'll propose those... | 15:47 |
| cdent | efried: the bytecode makes that much of a difference in the time to start the env? That's very surprising? | 15:49 |
| efried | cdent: deleting the bytecode could. Though placement is a pretty small repo, so I wouldn't expect it to be so bad. Worth trying. | 15:50 |
| cdent | i had assumed it was combo of installing/verifying the deps plus test discovery | 15:51 |
| efried | cdent: There was something else that was the main perf killer. I think it was something about ... yes, deps | 15:51 |
| efried | cdent: I believe there's a workaround where you can touch a file to make it not check the deps | 15:51 |
| efried | stephenfin: remind me please ^ | 15:51 |
| efried | cdent: for my part, I have some aliases set up that just run stestr directly. Because I got annoyed with all that and couldn't be arsed to keep remembering the workaround. | 15:52 |
| stephenfin | efried: touch nova.egg-info | 15:52 |
| efried | cdent: There ya go ^ | 15:52 |
| efried | thanks stephenfin | 15:52 |
| openstackgerrit | Eric Fried proposed openstack/placement master: tox: Don't write byte code (maybe) https://review.openstack.org/636995 | 15:55 |
| cdent | thanks efried and stephenfin | 15:56 |
| cdent | yeah, I often switch over to straight stestr or even unittest (depending on what I'm trying to do) | 15:56 |
| stephenfin | not that that shouldn't be necessary with the newest tox though, which does that exact thing itself | 15:56 |
| stephenfin | 3.1.3, apparently | 15:57 |
| stephenfin | (commit 95f5476e31fd20a93b70027efac1f1864b6ed25c) | 15:57 |
| cdent | but for the "just run all the things case" I feel obliged to do the defined thing and it gets me all narked that the test time is a tenth of a second but there's a minute of waiting | 15:57 |
| openstackgerrit | Eric Fried proposed openstack/placement master: Use tox 3.1.1 fixes https://review.openstack.org/636997 | 15:59 |
| efried | ah, neat stephenfin, good to know. Lemme go test that, since I done installed 3.7.0 yesterday... | 16:00 |
| efried | cdent: functional takes longer than a tenth of a second to run.... | 16:02 |
| cdent | efried: sure, but unit does not | 16:02 |
| efried | close to 15s on my machine | 16:02 |
| cdent | which is still pretty awesome, considering | 16:02 |
| efried | oh, fo sho compared to e.g. nova, which I don't think I've ever waited around for completion. | 16:02 |
| efried | cdent: new tox does indeed seem to get past the dep check way quicker. | 16:03 |
| efried | $ time tox | 16:04 |
| efried | <snip> | 16:04 |
| efried | py27: commands succeeded | 16:04 |
| efried | py36: commands succeeded | 16:04 |
| efried | functional: commands succeeded | 16:04 |
| efried | functional-py36: commands succeeded | 16:04 |
| efried | pep8: commands succeeded | 16:04 |
| efried | congratulations :) | 16:04 |
| efried | real1m6.563s | 16:04 |
| efried | user5m45.549s | 16:04 |
| efried | sys0m9.873s | 16:04 |
| cdent | yup, definitely speeds up dep check | 16:05 |
| cdent | so now it sits on stestr's discovery (fair enough) | 16:05 |
| cdent | thanks | 16:06 |
| efried | cdent: I haven't caught up on emails yet, so forgive me if already answered, but have you tested / seen any perf improvements from your nix-OVO efforts? | 16:11 |
| cdent | yeah, there's some converations in that in the thread about evaluating what was good or bad in placement, but the short version of the tale is maybe about 50% improvement | 16:12 |
| efried | noyce! | 16:12 |
| cdent | of course that's very handwavey imprecise | 16:12 |
| efried | Okay, but even with a massive margin of error it's an unquestionable improvement. | 16:12 |
| cdent | but the numbers go from 2.5ish to a bit more than 1.ish | 16:12 |
| cdent | yeah | 16:12 |
| efried | and we | 16:13 |
| cdent | there are some definite things to fix | 16:13 |
| efried | don't | 16:13 |
| efried | fin | 16:13 |
| efried | need | 16:13 |
| efried | ovo here | 16:13 |
| cdent | but i'd say it's prolly a goer | 16:13 |
| cdent | I mentioned on one of the patches that it might be useful for someone else to do an iteration on the stack | 16:13 |
| cdent | simply to get some different brain in there | 16:13 |
| cdent | my style tends to be "make the critical change and then adapt as tests fail" | 16:14 |
| cdent | which means things don't come out super clean (yet) | 16:14 |
| edleafe | Speed is nice, but removing a bunch of magic is even sweeter | 16:14 |
| cdent | yeah, that part is interesting too | 16:14 |
| cdent | there's been some implicit casting going on | 16:15 |
| cdent | from Decimal() to <other> which without ovo there leads to JSON serialization failures | 16:15 |
| cdent | (which was relatively easy to fix) | 16:15 |
| cdent | but the stuff that happens is now more clear | 16:16 |
| cdent | one thing that was extra clear while doing this work was that from a grepping standpoint, having the different objects in different files (or at least chunked) would have made this easier | 16:16 |
| cdent | (efried my time tox numbers (on hot venvs) is almost exactly the same) | 16:17 |
| cdent | (using new version) | 16:17 |
| efried | excellent | 16:39 |
| cdent | efried: your comments about deepcopy, remind me of a thing from the ovo removal stack. In the last one I specificaly draw your attention to some necessary deepcopy changes. | 16:55 |
| efried | looking... | 16:59 |
| efried | meanwhile - do you have any freakin idea how this could be happening? | 16:59 |
| efried | cdent: save me looking through all ten patches, where are said comments? | 17:00 |
| cdent | one mo | 17:00 |
| cdent | efried: end of https://review.openstack.org/#/c/636807/2/placement/objects/resource_provider.py two TODO with your nick | 17:01 |
| cdent | and the commit message | 17:01 |
| efried | why my nick? | 17:02 |
| efried | because I introduced the original deepcopyZ? | 17:02 |
| *** helenafm has quit IRC | 17:03 | |
| cdent | I think so, but I may have blamed wrong | 17:03 |
| efried | no, that looks like my code. | 17:03 |
| cdent | and by 1:30am last night I was shorthanding quite a bit to not forget stuff | 17:04 |
| efried | okay, duly noted, will investigate - but later. I'm holding up the bw series rn. | 17:04 |
| cdent | yeah, no rush, just didn't want to forget it | 17:05 |
| *** dims has quit IRC | 18:47 | |
| *** tssurya has quit IRC | 19:48 | |
| efried | cdent: What are you holding out on before opening up the OVO-nixing series to mergeability? | 20:15 |
| efried | The grenade and tempest things put it through live scenarios, right, so we're not just relying on ut/functional to be confident it's "right"? | 20:18 |
| *** e0ne has joined #openstack-placement | 20:21 | |
| cdent | efried: a) consensus that we should do it, b) SETTABLE_FIELDS (on rps) is currently disabled and needs to be reinstated in a slightly different form, c) (lower priority) there's a mess of FIXME throughout which are mostly to do with removing duplicatation and general cruft | 20:41 |
| cdent | but yes, grenade and tempest are making it do real things, and found a real problem (since fixed) | 20:42 |
| efried | cdent: ack. Finishing up real review of the bottom one. | 20:42 |
| cdent | kewl | 20:43 |
| efried | cdent: done | 20:44 |
| cdent | thanks, those make sense, mostly | 20:46 |
| efried | cdent: added one followon to the usage thing. | 20:47 |
| * cdent nods | 20:47 | |
| efried | cdent: if a), and b) is later in the series, and c) n/a for the bottom patch (which it will be if you go forward with my comments), can you unblock? | 20:49 |
| *** takashin has joined #openstack-placement | 20:52 | |
| cdent | How could (a "get consensus") be later in the series. Presumably we either de-OVO or don't, not a hybrid? | 20:54 |
| cdent | b is later in the series. | 20:54 |
| cdent | What is your goal in unblocking asap? | 20:54 |
| cdent | efried: responded, but mostly I agree with your comments | 21:00 |
| efried | cdent: reword: Given that (a) is true (i.e. we have consensus); and (b) happens later in the series; ... | 21:00 |
| efried | Goal in unblocking fast, nothing really, just would rather it got attention while it's fresh in people's minds. With a -2 people will probably be hesitant to render real votes, and then it'll wind up languishing and not getting finished. | 21:01 |
| efried | And this is something I've been wanting to see done for quite a while, so I'm mainly just excited to see it happen. | 21:02 |
| cdent | reckon so, depends on how quickly there's some constructive review on the later bits of the series. There's some commonality in all the *List classes that could be done from the start, or evolve as we go. I don't really have a preference. Probably just being shy because of nova ptsd. | 21:02 |
| cdent | what would be most useful to me, from you, is rather than a close review of each patch in series is a superficial review of the whole thing, first | 21:03 |
| cdent | then go back for the closer review | 21:03 |
| cdent | but I agree, yes, it would be good to keep the ball rolling. assuming there's no "omg, no" before my EOD tomorrow I'll make sure the earliest patches are unblocked | 21:04 |
| efried | cdent: well, my superficial review of the whole thing (as it stood at the time) was what yielded "Huge ++ to this idea, Chris." | 21:14 |
| efried | And my comments on the bottom change's commit message go a bit deeper in support of the thing overall. | 21:15 |
| cdent | efried: sure, but right now, though all tests pass it does not have identical behavior to master, especially since at least two tests are disabled | 21:15 |
| efried | So I'm not sure there's value in a less-superficial-but-still-not-thorough review | 21:15 |
| cdent | i'm after things like "I think you should make List super class" and "there needs to be a single _updates method" | 21:16 |
| efried | noted | 21:25 |
| -openstackstatus- NOTICE: Jobs are failing due to ssh host key mismatches caused by duplicate IPs in a test cloud region. We are disabling the region and will let you know when jobs can be rechecked. | 21:32 | |
| *** e0ne has quit IRC | 21:48 | |
| cdent | night all | 21:57 |
| *** cdent has quit IRC | 21:57 | |
| -openstackstatus- NOTICE: The test cloud region using duplicate IPs has been removed from nodepool. Jobs can be rechecked now. | 22:14 | |
| openstackgerrit | Merged openstack/placement master: Trivial: return empty set instead of list https://review.openstack.org/636846 | 23:07 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!