| *** lemko has joined #openstack-loci | 05:58 | |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: Always build requirements image in the pipeline https://review.openstack.org/609646 | 06:49 |
|---|---|---|
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: Always build requirements image in the pipeline https://review.openstack.org/609646 | 06:51 |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: Always build requirements image in the pipeline https://review.openstack.org/609646 | 06:51 |
| evrardjp | good morning | 06:55 |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: Always build requirements image in the pipeline https://review.openstack.org/609646 | 07:55 |
| *** e0ne has joined #openstack-loci | 07:59 | |
| *** e0ne has quit IRC | 08:44 | |
| *** e0ne has joined #openstack-loci | 08:46 | |
| hrw | https://marcin.juszkiewicz.com.pl/2018/10/12/how-fast-is-apm-mustang/ shows how fast my arm64 devbox is | 08:51 |
| evrardjp | hrw: thanks for the vote :) | 08:57 |
| evrardjp | hrw: haha | 08:58 |
| evrardjp | hrw: good to see a comparison. I sometimes see ppl comparing cores as equals where they are effectively comparing apple to oranges | 08:59 |
| hrw | yep | 09:00 |
| hrw | hm. I have access to 2x48 cores machine... | 09:00 |
| hrw | good that I logged in. 7.7GB left of storage... | 09:02 |
| *** e0ne has quit IRC | 09:13 | |
| *** e0ne has joined #openstack-loci | 09:16 | |
| *** e0ne has quit IRC | 10:34 | |
| hrw | what do you think about adding squashing to loci images? | 11:08 |
| *** e0ne has joined #openstack-loci | 11:27 | |
| *** e0ne has quit IRC | 12:16 | |
| *** e0ne has joined #openstack-loci | 12:17 | |
| *** munimeha1 has joined #openstack-loci | 13:22 | |
| *** munimeha1 has quit IRC | 13:38 | |
| *** e0ne has quit IRC | 14:58 | |
| evrardjp | I trust you hrw :) | 15:00 |
| evrardjp | but... there is a meeting today, and I'd like to chat with people about what I did recently and what's pending | 15:01 |
| evrardjp | so maybe it could be something we can talk about too? | 15:01 |
| *** e0ne has joined #openstack-loci | 15:06 | |
| evrardjp | is there someone here for the meeting? | 15:28 |
| evrardjp | cc: portdirect pbourke hogepodge ^ | 15:29 |
| *** e0ne has quit IRC | 15:37 | |
| hogepodge | Hi | 15:38 |
| evrardjp | hello | 15:38 |
| *** e0ne has joined #openstack-loci | 15:40 | |
| evrardjp | I'd hope we have more than 2 or 3 pairs of eyes for this meeting, do we have a quorum? | 15:40 |
| hogepodge | We need to get the project back on track. portdirect and samyaple can we confirm for next week? | 15:40 |
| evrardjp | I have plenty of things pending there | 15:40 |
| *** e0ne has quit IRC | 15:40 | |
| evrardjp | At some point we can't wait anymore | 15:40 |
| evrardjp | (at least I have employer pressure there) | 15:41 |
| hogepodge | reviewing | 15:46 |
| evrardjp | thanks hogepodge | 15:47 |
| hogepodge | evrardjp: I'm very reluctant to merge https://review.openstack.org/#/c/602186/5 | 15:52 |
| hogepodge | I think that adding project-specific bindep files is a bad idea. | 15:52 |
| evrardjp | the previous iteration which was including a full bindep copy was approved by sam | 15:52 |
| evrardjp | I thought this was already better | 15:53 |
| evrardjp | but I am open to discussion ofc :) | 15:53 |
| evrardjp | see https://review.openstack.org/#/c/577667/ | 15:53 |
| hrw | re | 15:53 |
| hogepodge | I like this one https://review.openstack.org/#/c/602180/ | 15:54 |
| hogepodge | Yeah, I see I +2ed that also. | 15:55 |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/loci master: Introduce extra *dep parameter https://review.openstack.org/602180 | 15:55 |
| evrardjp | yeah the extra *dep file was used as basis to have OSH image building basically | 15:56 |
| evrardjp | the limitation of Docker in terms of file path makes it easier to carry OSH file into a subfolder of loci vs being external, as far as I am told | 15:57 |
| hogepodge | Would https://review.openstack.org/#/c/602180/6 be sufficient for you? I've really refined my views and don't want to host project-specific enhancements for a whole bunch of reasons. | 15:58 |
| evrardjp | external and on the filesystem I mean | 15:58 |
| hogepodge | Kubernetes cloud provider work has really opened my eyes to how bad things can get once you open the door to it. | 15:58 |
| evrardjp | hogepodge: I think it's worth being explicit | 15:58 |
| evrardjp | could you tell us more? | 15:58 |
| evrardjp | I think I kinda agree with you | 15:58 |
| evrardjp | but I'd rather have it documented somewhere -- and we should explain the reasons why | 15:59 |
| hogepodge | * There's no guarantee that any project will live longer than Loci. If that happens, who becomes responsible for removing that code. What's the condition or trigger for it? | 15:59 |
| evrardjp | I mean the trend was, currently, to have kolla and OSH images inside LOCI umbrella. | 15:59 |
| evrardjp | loci code is still managed by loci team | 16:00 |
| evrardjp | it can arbitrary remove or add code | 16:00 |
| hogepodge | * What's the criteria for allowing an external project to store their configuration in a general repository? How do you guarantee fairness? | 16:00 |
| evrardjp | that was my concern | 16:00 |
| evrardjp | :D | 16:00 |
| evrardjp | you hit it! ;) | 16:00 |
| hogepodge | * If there's a critical bug in project-specific files that impacts an external project, how do we get quick review in on a project they don't own? | 16:01 |
| evrardjp | I think with the fact that we can chain bindeps/pydeps we could technically have project jobs that are building images based on LOCI images, by having a pre-job creating a webserver to publish said bindep/pydep file, and point to those in extra_* | 16:01 |
| hogepodge | Having a pluggable mechanism solves every one of those potential problems. | 16:01 |
| evrardjp | That's why I implemented both into separate commits :D | 16:02 |
| evrardjp | (btw I have updated 602180 | 16:02 |
| evrardjp | hogepodge: I think it's fair to document this in the README that LOCI won't ship extra bindep/pydep files | 16:03 |
| hogepodge | +1 to that. I can still be overridden by other cores, but that's my stance on it right now. | 16:04 |
| evrardjp | I am not core but my experience tells me it's a better path. I proposed the other patch due to willingness of others, and my willingness to advance :) | 16:05 |
| evrardjp | better advance and fail, than standstill | 16:05 |
| evrardjp | anyway | 16:05 |
| evrardjp | I have two other topics I'd like to discuss | 16:05 |
| evrardjp | https://review.openstack.org/#/c/609646/ and https://review.openstack.org/#/c/609616/ | 16:05 |
| hogepodge | evrardjp: I hope you don't think I'm beating up on the patches. | 16:06 |
| hogepodge | I definitely should have been reviewing earlier also. | 16:06 |
| evrardjp | hogepodge: not sure what you meant there. Patches are meant for review? :) | 16:07 |
| evrardjp | I have those kinda of ideas in my head for a while, but we didn't get the chance to discuss this. Now I need to discuss it and go further | 16:07 |
| evrardjp | :D | 16:07 |
| evrardjp | so those two patches are important for a contributor named Iago Santos | 16:08 |
| evrardjp | it would also be usable by other consuming projects. | 16:08 |
| hogepodge | So on the first, I don't fully understand why we were caching requirements. I wish SamYaple were here to talk a bit more about it. | 16:08 |
| hogepodge | I'm going to ask in the infra room because I don't want to hit them with something that may but burden on them | 16:09 |
| evrardjp | portdirect: maybe you have more background there? | 16:09 |
| hrw | hogepodge: what you by 'we were caching requirements'? | 16:09 |
| hrw | s/you/you mean/ | 16:09 |
| evrardjp | I don't believe my reasoning to be false though, but I am willing to see if there are constructive ways to do things in zuul that would have the best of both (fan out jobs?) | 16:10 |
| evrardjp | hrw: see l23 to 46 on the right there: https://review.openstack.org/#/c/609646/8/playbooks/loci-builder.yaml | 16:10 |
| evrardjp | on the right side* | 16:10 |
| hrw | that makes sense | 16:11 |
| hrw | as 'that change' | 16:11 |
| hrw | when we want to build new set of images we get whole set | 16:11 |
| hrw | and on x86 building requirements takes moments | 16:12 |
| hogepodge | https://github.com/openstack/loci/blob/2df96283b9cdd8b6934d1f8321700281bedeeaeb/playbooks/loci-builder.yaml#L13 | 16:12 |
| evrardjp | the other patch brings branching, which means we could create branch job for every project. I like the idea of building multiple branches with the same codebase, and have these jobs. | 16:13 |
| evrardjp | hogepodge: I didn't remove said feature. I just don't believe it's the right thing to activate by default in checks/gates. | 16:13 |
| hogepodge | evrardjp: right, I just want to make sure infra is good with the addition workload and bandwidth | 16:14 |
| evrardjp | should someone want to reuse he can still pass the flag. It makes sense for example to pass the flag to EOL branches or EM branches. | 16:14 |
| evrardjp | hogepodge: yes, and maybe constructive ideas on how to build the jobs in a fanout fashion | 16:15 |
| evrardjp | hogepodge: would you be okay for adding jobs per (project,branch) ? | 16:15 |
| evrardjp | also if I can get an extra vote on https://review.openstack.org/#/c/608287/, that would be cool | 16:17 |
| evrardjp | that's all I had for this meeting | 16:17 |
| *** irclogbot_3 has joined #openstack-loci | 16:35 | |
| *** irclogbot_3 has quit IRC | 16:42 | |
| openstackgerrit | Merged openstack/loci master: Add 'opensuse' key in distro check https://review.openstack.org/608287 | 16:52 |
| *** e0ne has joined #openstack-loci | 17:28 | |
| openstackgerrit | Merged openstack/loci master: Introduce extra *dep parameter https://review.openstack.org/602180 | 18:05 |
| *** e0ne has quit IRC | 18:37 | |
| *** kmalloc is now known as needscoffee | 18:48 | |
| *** needscoffee is now known as kmalloc | 20:20 | |
| *** lemko has quit IRC | 20:58 | |
| *** pbourke has quit IRC | 22:32 | |
| *** pbourke has joined #openstack-loci | 22:32 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!