| *** aostapenko has quit IRC | 01:44 | |
| *** openstackgerrit has joined #openstack-loci | 04:47 | |
| openstackgerrit | Merged openstack/loci master: Improve setup-gate https://review.openstack.org/609623 | 04:47 |
|---|---|---|
| *** e0ne has joined #openstack-loci | 06:37 | |
| evrardjp | I cannot attend today's meeting | 07:05 |
| evrardjp | hrw: are you there by any chance? | 07:05 |
| hrw | yep | 07:06 |
| hrw | what time meeting is? | 07:06 |
| evrardjp | 3:30 UTC IIRC | 07:08 |
| evrardjp | sorry | 07:08 |
| evrardjp | 3:00 UTC | 07:08 |
| hrw | 17:00 my time. will add to calendar as I am forgetting about it | 07:08 |
| evrardjp | hrw: I am stuck on something that seem to work on my machine. Could you lend me some eyes? | 07:09 |
| evrardjp | http://logs.openstack.org/03/611303/12/check/loci-nova/b6602a3/ara-report/result/ff59c32d-1c51-4e6d-b759-c74adfa8b1f7/ last line | 07:09 |
| evrardjp | I changed fetch_wheels.sh to show what's going on : https://review.openstack.org/#/c/611303/12/scripts/fetch_wheels.sh | 07:09 |
| hrw | line 18: gunzip: command not found | 07:10 |
| evrardjp | so it seems on my image gunzip is not there, fine, but tar is obviously found | 07:10 |
| evrardjp | the file is found too | 07:10 |
| evrardjp | (did on another patchset a ls of /tmp/ and a file of /tmp/wheels.tar.gz ) | 07:10 |
| evrardjp | file was telling me it was a compressed archive | 07:11 |
| evrardjp | so I guess MIME type is correct | 07:11 |
| hrw | evrardjp: do not give compression type to tar when you are unpacking | 07:11 |
| hrw | 'tar xf' is enough | 07:11 |
| hrw | unless you are on 2011 distro | 07:11 |
| evrardjp | that's what's done on l23 | 07:11 |
| evrardjp | still failing :p | 07:12 |
| hrw | yes. I spoke about l17 | 07:12 |
| evrardjp | I wanted to try to be explicit to see if it worked better. as you see, it doesn't. | 07:12 |
| evrardjp | So I am curious what could be going on | 07:12 |
| evrardjp | 1) I am thinking of apparmor but it works perfectly fine on my machine | 07:13 |
| hrw | evrardjp: 'zypper install /usr/bin/gzip' or sth like that | 07:13 |
| evrardjp | gzip is installed by default | 07:13 |
| evrardjp | it's required for zypper | 07:13 |
| evrardjp | I can be explicit | 07:13 |
| hrw | add 'which gzip' 'which gunzip' to job and retry? | 07:13 |
| hrw | as it cleanly misses those | 07:14 |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: [WIP] Build OpenSUSE Leap 15 by default https://review.openstack.org/611303 | 07:14 |
| evrardjp | which is not available but I can test | 07:14 |
| hrw | install gzip explicitly? | 07:15 |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: [WIP] Build OpenSUSE Leap 15 by default https://review.openstack.org/611303 | 07:16 |
| evrardjp | hrw: I just pushed a patch with those things | 07:17 |
| hrw | will see how it goes | 07:17 |
| evrardjp | hrw: I think it would be beneficial for me to get the archive in the logs | 07:17 |
| hrw | ;D | 07:17 |
| * hrw fights with pbuilder on debian | 07:18 | |
| evrardjp | hrw: do you know the way to export a docker image? I'd be curious to see what's inside said image after its failure | 07:20 |
| hrw | evrardjp: why not build that image on your machine? | 07:20 |
| evrardjp | it works on my machine : ) | 07:21 |
| evrardjp | I am building a requiremnt container just fine, and using said container to build the other images just fine | 07:21 |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: [WIP] Build OpenSUSE Leap 15 by default https://review.openstack.org/611303 | 07:38 |
| evrardjp | if it doesn't work I am abandonning. | 07:38 |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: [WIP] Build OpenSUSE Leap 15 by default https://review.openstack.org/611303 | 07:41 |
| *** e0ne has quit IRC | 08:03 | |
| evrardjp | hrw: the plot thickens: http://logs.openstack.org/03/611303/16/check/loci-cinder/f0ca917/ara-report/ | 08:45 |
| evrardjp | even the save of the archive is doing weird stuff | 08:45 |
| evrardjp | I mean the docker save output has failure to be fetched into gates | 08:45 |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: [WIP] Build OpenSUSE Leap 15 by default https://review.openstack.org/611303 | 08:53 |
| *** e0ne has joined #openstack-loci | 09:09 | |
| *** lemko has joined #openstack-loci | 10:52 | |
| *** e0ne has quit IRC | 11:01 | |
| *** e0ne has joined #openstack-loci | 11:52 | |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: [WIP] Build OpenSUSE Leap 15 by default https://review.openstack.org/611303 | 11:54 |
| *** munimeha1 has joined #openstack-loci | 13:53 | |
| *** lemko has quit IRC | 14:02 | |
| openstackgerrit | Andrii Ostapenko proposed openstack/loci master: Install pip packages during requirements build https://review.openstack.org/611451 | 14:26 |
| hrw | I wonder how much that change will add to build time | 14:42 |
| *** e0ne has quit IRC | 14:52 | |
| hogepodge | Build size also | 14:59 |
| hogepodge | Anyone here for the Loci meeting? | 15:00 |
| hrw | o/ | 15:04 |
| hrw | hogepodge: on !x86 build time is more important. storage is cheap | 15:04 |
| hogepodge | Build size is an important design consideration in this project | 15:05 |
| hogepodge | As is time and correctness | 15:06 |
| hrw | hogepodge: sure | 15:06 |
| hrw | hogepodge: but requirements image does not land on target systems | 15:06 |
| hrw | it is used only to build deployed images | 15:06 |
| hrw | so it does not affect setups much | 15:06 |
| hogepodge | Regardless, we can look at the time metrics from the gate job and compare without the patch. Presumably it’s installing locally as a cache for the wheel build? If it’s not doing that it’s not of much use | 15:11 |
| hogepodge | It’s meant to pin upper constraints | 15:11 |
| hrw | I feel that it can be a way for workaround some issue | 15:11 |
| hrw | 23:58 < aostapenko> Hi all, we're using loci to build images for Ocata and faced https://github.com/lxc/pylxd/issues/308 during creating wheel for pylxd 2.2.4 during requirements image build. As 2.2.4 is a pinned version for Ocata and bug is only fixed in 2.2.7, we're thinking to work this around with installing urllib3 prior to wheel build. Apparently we would like to have desired functionality in | 15:12 |
| hrw | upstream, as we directly use upstream loci i | 15:12 |
| hrw | 23:58 < aostapenko> scripts on fly. I've prepared a change https://review.openstack.org/#/c/611451/ that adds an ability to work around such issues, may you please review it. Probably in this particular case it won't break anything if pylxd with this issue fixed is installed instead of 2.2.4, however I consider this as a bad solution. | 15:12 |
| hrw | yep | 15:12 |
| hogepodge | We can include a list of problematic libraries and do a local pip install of those. It’s a better solution | 15:16 |
| hrw | can you comment in patch review? | 15:16 |
| hogepodge | Of course, just using the meeting time for discussion :-) | 15:17 |
| hrw | sure ;) | 15:17 |
| hrw | started build of requirements on arm64. will later do with patch applied | 15:21 |
| hogepodge | It could also be another configuration option, but it would be another config option | 15:21 |
| hrw | btw - is there a way to tell pip to not generate "Ignoring typed-ast: markers 'python_version == "3.4"' don't match your environment" like messages? | 15:22 |
| hogepodge | Probably some use of the -q flag, which can be applied repeatedly to suppress messages | 15:24 |
| hrw | thx | 15:27 |
| hogepodge | Just looking at the gate, it looks like the patch increased build time by 40%. | 15:29 |
| hogepodge | That could be cloud dependent though, so a local test is in order (when you get numbers that would be useful for the review). As it stands I think the patch is too crude of a solution to land the way it is. If it didn't impact time so significantly I'd probably let it in since it would just be acting as a cache. | 15:31 |
| hrw | yep | 15:31 |
| hogepodge | Were there any other issues we wanted to cover this in this week's highly informal meeting? | 15:34 |
| hrw | evrardjp fights with his opensuse patch: https://review.openstack.org/611303 | 15:37 |
| hrw | for me it looked as image issue but it works for him on local machine | 15:38 |
| hogepodge | Build logs weren't copied over. | 15:40 |
| hogepodge | The build went fine for requirements, but when logs failed is posted a -1. | 15:40 |
| hogepodge | Same for all the other image builds. Might want to have infra take a look | 15:41 |
| hrw | evrardjp: ^^ | 15:42 |
| hrw | that's all I think | 15:42 |
| hogepodge | I just left a comment on the patch | 15:46 |
| hogepodge | evrardjp: don't be afraid to reach out to infra for help on creating a new build job. They like it when people reach out | 15:47 |
| hogepodge | that's it for me, I'll leave a review on the other patch | 15:47 |
| hrw | thanks for meeting | 15:47 |
| *** openstackgerrit has quit IRC | 16:24 | |
| *** e0ne has joined #openstack-loci | 16:35 | |
| *** e0ne has quit IRC | 16:37 | |
| hogepodge | I'll use startmeeting next time so it's properly logged | 16:55 |
| hogepodge | hrw I'm not necessarily seeing significant changes in build time on x86 | 17:32 |
| *** e0ne has joined #openstack-loci | 17:53 | |
| *** e0ne has quit IRC | 18:00 | |
| *** e0ne has joined #openstack-loci | 18:08 | |
| hrw | on arm64 it's 42-43 minutes each time so no change too | 19:01 |
| *** e0ne has quit IRC | 19:34 | |
| *** e0ne has joined #openstack-loci | 19:34 | |
| *** e0ne has quit IRC | 19:36 | |
| *** e0ne has joined #openstack-loci | 20:28 | |
| *** e0ne has quit IRC | 20:33 | |
| *** e0ne has joined #openstack-loci | 20:44 | |
| *** e0ne has quit IRC | 20:48 | |
| *** munimeha1 has quit IRC | 20:51 | |
| *** pbourke has quit IRC | 22:36 | |
| *** pbourke has joined #openstack-loci | 22:37 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!