| *** lemko has quit IRC | 00:47 | |
| *** e0ne has joined #openstack-loci | 06:18 | |
| *** e0ne has quit IRC | 06:43 | |
| *** e0ne has joined #openstack-loci | 06:46 | |
| *** e0ne has quit IRC | 06:48 | |
| *** lemko has joined #openstack-loci | 07:02 | |
| *** e0ne has joined #openstack-loci | 07:05 | |
| *** e0ne has quit IRC | 07:07 | |
| evrardjp | hogepodge: worked for me yesterday in my build env | 10:53 |
|---|---|---|
| *** e0ne has joined #openstack-loci | 12:47 | |
| *** lemko has quit IRC | 13:15 | |
| *** e0ne has quit IRC | 13:28 | |
| *** e0ne has joined #openstack-loci | 13:30 | |
| *** mandre is now known as mandre_away | 14:02 | |
| *** lemko has joined #openstack-loci | 14:14 | |
| *** e0ne has quit IRC | 14:33 | |
| hogepodge | Hi everyone | 15:00 |
| hogepodge | Meeting today? | 15:02 |
| hrw | hi | 15:02 |
| hrw | anything to discuss? | 15:03 |
| hrw | https://review.openstack.org/632129 makes me sad as I do not want to rewrite how requirements are built | 15:03 |
| hrw | as current implementation is based on luck | 15:04 |
| hogepodge | Yeah, figuring out a solution for ARM is important. | 15:05 |
| hogepodge | I have to admit I don't completely understand the problem, but it sounds like part of it is a race condition in the build? | 15:06 |
| hrw | like I said - it is luck based process | 15:06 |
| *** e0ne has joined #openstack-loci | 15:07 | |
| hrw | it runs 'pip install onepackage' hoping that there will be no deps installed | 15:07 |
| hrw | and once they are whole process goes into wild | 15:07 |
| hogepodge | So rather than remove the pruning step we should start by doing a serial build and remove the -P20 from 48, then see which packages are problematic | 15:07 |
| hrw | any package can be problematic | 15:08 |
| hogepodge | why is --no-deps not working on pip wheel? | 15:08 |
| hrw | hogepodge: because some packages need to be built | 15:08 |
| hrw | on x86 you fetch binaries from pypi | 15:09 |
| hrw | on !x86 you build. even when want wheel | 15:09 |
| hogepodge | Also, if OpenStack has dependencies that aren't defined that's a but in requirements that's an openstack bug | 15:09 |
| hogepodge | s/but/bug/ | 15:10 |
| hogepodge | sorry for my early morning sentence mangling. | 15:10 |
| hrw | no problem | 15:11 |
| hrw | 16:11 here | 15:11 |
| evrardjp | hey | 15:11 |
| hogepodge | requirements is supposed to capture the dependency tree and give operators and packagers a list of things they need, which is why we have it | 15:11 |
| hrw | https://github.com/openstack/requirements/blob/master/global-requirements.txt lists cffi | 15:11 |
| hrw | and loci/requirements builds cffi twice | 15:11 |
| evrardjp | the problem is we don't do a difference for a skip (do not build) and a skip (already build), correct? | 15:12 |
| hrw | yes | 15:12 |
| evrardjp | it sounds like a us problem | 15:13 |
| evrardjp | 'us' | 15:13 |
| hogepodge | then the Skipping search criteria is a bad one, so that's where we should focus then. That line is a bug, and removing it isn't the answer. | 15:13 |
| hrw | relying on pip output suxx | 15:13 |
| evrardjp | hogepodge: agreed | 15:13 |
| hrw | we may drop *none.whl files probably as they will be fetched anyway | 15:13 |
| hrw | and keep all architecture related ones - nevermind are they built of fetched | 15:14 |
| evrardjp | sounds reasonable | 15:14 |
| hogepodge | loci is a giant pile of hacks | 15:14 |
| evrardjp | for sure | 15:14 |
| hrw | personally I would just ignore 'omg, image has grown by 20%' rule as if someone has a problem with extra 60MBxfew then they have other issues than bigger image imho | 15:15 |
| evrardjp | I think it would be neater if we remove some of those, but loose efficiency | 15:15 |
| hogepodge | we could reduce the size to zero and the build would still be incorrect, and we could all go home ;-) | 15:15 |
| hrw | evrardjp: I am step from adding 'if !x86' shell check around that line and going with devel | 15:15 |
| hrw | or just a line into CI job which will drop that line ;D | 15:16 |
| hogepodge | that is to say, size doesn't matter if we can't do the right thing, and I'd rather do the right thing | 15:16 |
| hogepodge | My proposal is to drop in a temporary fix that we know works for ARM build like you're suggesting so that you're not blocked, but look as a correct fix that balances our requirements for minimization. | 15:17 |
| hrw | imho process can be changed to "build requirements.tar, do builds with mounting local dir into containers" | 15:18 |
| hrw | as fetching 260MB file for each image sounds even worse | 15:18 |
| hogepodge | We need better documentation and tooling for handling requirements offline | 15:19 |
| hogepodge | It's possible to cache requirements locally and just connect to a file, but the default is to pull. | 15:19 |
| hogepodge | I feel the pain on downloading requirements, I've gone past my data cap twice working on Loci because of downloading requirements over and over. | 15:20 |
| hrw | hogepodge: you live in Canada or USA? | 15:21 |
| hogepodge | USA | 15:21 |
| hrw | data cap. | 15:21 |
| evrardjp | hrw I like that idea myself (the folder), or have a multiple step build | 15:21 |
| evrardjp | (targets) | 15:21 |
| hrw | Canada and USA are 4th world countries when it comes to internet | 15:21 |
| hogepodge | #freedom ;-) | 15:22 |
| hrw | whenever I hear 'I have data cap' it is US citizen. | 15:22 |
| hrw | or we talk about mobile and cap means 'just 10GB per month' | 15:22 |
| * hrw hugs 2.3TB data cap on mobile | 15:23 | |
| evrardjp | wow we cap at 2oGB on mobile here. | 15:23 |
| evrardjp | 20* | 15:23 |
| hrw | evrardjp: I pay 25 pln (~6€) monthly for unlimited calls/texts and 30+GB of data. Data cumulates. | 15:24 |
| evrardjp | nice | 15:24 |
| hogepodge | hrw: so how do you feel about the proposed plan, a hack to fix the ARM build along with a plan for a correct evaluation of wheels | 15:24 |
| hrw | so going back to our pile of hacks called loci... | 15:24 |
| hrw | hogepodge: sounds good. | 15:24 |
| evrardjp | question | 15:24 |
| evrardjp | hrw: will you be at PTG? Should we have a hacking session? | 15:25 |
| hrw | evrardjp: nope | 15:25 |
| evrardjp | would that help on your arm story? | 15:25 |
| evrardjp | ok | 15:25 |
| hogepodge | I have a half day scheduled for Loci at PTG in Denver | 15:25 |
| evrardjp | k | 15:25 |
| hogepodge | hrw: if you can get me ARM hardware access I'm happy to look at solutions also. | 15:26 |
| evrardjp | hogepodge: :D | 15:26 |
| hrw | hogepodge: ;D will ask | 15:26 |
| hrw | I may have some resources on linaro cloud | 15:28 |
| hogepodge | I may also, I think gema set up an account for me a while ago now that I think about it. | 15:28 |
| hrw | voila! | 15:29 |
| hrw | looked. no resources for me there | 15:29 |
| hogepodge | I'll check with her and follow up | 15:29 |
| hrw | cool | 15:29 |
| hogepodge | speaking of hacks, I'm inclined to merge this so that evrardjp isn't blocked any more either. https://review.openstack.org/#/c/624396/ | 15:31 |
| hrw | fine for me | 15:32 |
| evrardjp | haha yeah -- I think we should have blacklists | 15:32 |
| hrw | qpid was always source of issues to me | 15:32 |
| evrardjp | what openstack branch do you build hrw? | 15:32 |
| *** kmalloc is now known as needscoffee | 15:32 | |
| hrw | evrardjp: master only | 15:34 |
| hrw | I left stable branches for others | 15:34 |
| hogepodge | in my experience master is the only reliable build | 15:34 |
| evrardjp | hogepodge: true | 15:35 |
| hogepodge | so many of the hacks are because of limitations set by packagers | 15:35 |
| evrardjp | sad but true | 15:35 |
| hogepodge | for example, once RH ships with a non-broken virtualenv we can revisit this https://review.openstack.org/#/c/636447/ | 15:36 |
| evrardjp | oh yeah it works for me in my env | 15:38 |
| evrardjp | so I am fine for that | 15:38 |
| hogepodge | Unrelated, I want to send a patch up that removes the Horizon build. Because it's a web app it's installed in a completely different way, and the Horizon loci build is unusable | 15:38 |
| hogepodge | which shines a light on the need for functional testing | 15:38 |
| hogepodge | I've been building this for a simple way to do an AIO installation with Loci, which once I set up testing I may move to openstack hosting and gate against. The downside is it's only CentOS because of sunk costs and limited hours in the day. https://github.com/hogepodge/locistack | 15:39 |
| hrw | I need to reinit whole local loci :( | 15:40 |
| hrw | 5 commits in repo | 15:40 |
| evrardjp | hogepodge: I would be fine removing horizon | 15:45 |
| evrardjp | I have no use for it | 15:45 |
| evrardjp | well I kinda do, but it's low prio for me | 15:45 |
| hogepodge | someone will, we just need to special case it like we do requirements. It makes no sense to gate against something that doesn't work | 15:45 |
| hogepodge | We'll ruin the infra data cap :-D | 15:45 |
| evrardjp | I think osh will have to adapt to that decision | 15:46 |
| evrardjp | I think for now I would be fine removing it from the voting with no doubt | 15:47 |
| evrardjp | hogepodge: could you clarify on the "installed in a completely different way" ? | 15:48 |
| hogepodge | evrardjp: pip install doesn't do the job | 15:48 |
| evrardjp | you compare to services behind uwsgi because django? | 15:48 |
| evrardjp | oh it's not enough you mean | 15:49 |
| evrardjp | ok | 15:49 |
| evrardjp | (didn't touch horizon for years) | 15:49 |
| hogepodge | It does pip install, but there's so much more that you have to do to httpd https://docs.openstack.org/horizon/latest/install/from-source.html | 15:49 |
| evrardjp | yes I am not surprised | 15:50 |
| hogepodge | As far as I understand it for source install you need the horizon repo manage.py to generate a bunch of stuff for you. https://docs.openstack.org/horizon/latest/install/from-source.html#deployment | 15:50 |
| hogepodge | I started working on in then decided Cinder was more important to me | 15:50 |
| hogepodge | Anything else in the last few minutes? I have another meeting starting soon. | 15:52 |
| * hrw done | 15:54 | |
| hogepodge | thanks everyone. hrw if you add that boundary around the patch we can merge. Also, can you send me a paste of the ARM build output so I can start thinking about a better way to do the package install pruning? | 15:54 |
| hogepodge | thanks hrw and evrardjp | 15:56 |
| hogepodge | we should also just mention portdirect because it's fun to do that ;-) | 15:56 |
| evrardjp | haha | 16:00 |
| evrardjp | "what happened"? | 16:00 |
| *** openstackgerrit has joined #openstack-loci | 16:17 | |
| openstackgerrit | Merged openstack/loci master: Remove python-qpid-proton 0.14.0 build https://review.openstack.org/624396 | 16:17 |
| *** lemko has quit IRC | 16:25 | |
| *** dims has joined #openstack-loci | 16:34 | |
| *** dims has quit IRC | 16:38 | |
| *** dims has joined #openstack-loci | 16:45 | |
| *** dims has quit IRC | 16:53 | |
| *** dims has joined #openstack-loci | 16:55 | |
| *** e0ne has quit IRC | 16:59 | |
| evrardjp | hogepodge: I think we should have startmeeting and endmeeting to have the minutes posted | 17:03 |
| hogepodge | evrardjp: yeah, I wasn't sure if the meeting was going to happen, then things just took off. definitely my mistake. | 17:04 |
| evrardjp | hogepodge: don't beat yourself for it :p | 17:07 |
| evrardjp | it's fine | 17:07 |
| openstackgerrit | Merged openstack/loci master: Fix broken symlinks from virtualenv 16.4.0 update https://review.openstack.org/636447 | 17:42 |
| *** lemko has joined #openstack-loci | 18:22 | |
| *** e0ne has joined #openstack-loci | 19:19 | |
| -openstackstatus- NOTICE: The StoryBoard service on storyboard.openstack.org is offline momentarily for maintenance: http://lists.openstack.org/pipermail/openstack-discuss/2019-February/002666.html | 20:03 | |
| *** e0ne has quit IRC | 20:51 | |
| *** lemko has quit IRC | 23:12 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!