| clarkb | SpamapS: my hunch is that the vast amount of io that ansible and git do is improved when jumping between page tables | 00:00 |
|---|---|---|
| clarkb | but I haven't really dug in because I'm happy to take the performance bump as is :) | 00:00 |
| fungi | a gift from the kernel gods | 00:00 |
| clarkb | interestingly this improvment appears to come without pcid flag on our VM cpus | 00:02 |
| openstackgerrit | Jeremy Stanley proposed openstack-infra/zuul-website master: Add a temporary summit promotional message https://review.openstack.org/591870 | 00:05 |
| fungi | that's very much wip | 00:05 |
| *** rlandy has quit IRC | 00:43 | |
| openstackgerrit | Jeremy Stanley proposed openstack-infra/zuul-website master: Add a temporary summit promotional message https://review.openstack.org/591870 | 01:27 |
| fungi | i added a block to the css to override styling of p in a banner, but kept consistent with the existing color scheme | 01:30 |
| corvus | fungi: cool, that generally looks good.... but now that i see that, perhaps we should also post an ansiblefest thing -- after all, we'll have a booth there | 01:59 |
| fungi | indeed! | 01:59 |
| corvus | fungi: you might consider a little more horizontal space between the text and the button, and a slightly larger font on the button | 02:00 |
| fungi | heh, jimmy just suggested that in another conversation (the extra space). i had it in locally and removed it because i wasn't sure it was warranted, but easy to add | 02:00 |
| corvus | good typography is universal! :) | 02:01 |
| fungi | and if i just switch to the normal button class instead of the small button, that probably addresses the size concern | 02:01 |
| openstackgerrit | Jeremy Stanley proposed openstack-infra/zuul-website master: Add a temporary summit promotional message https://review.openstack.org/591870 | 02:13 |
| fungi | see if that's ^ any better | 02:13 |
| *** openstack has joined #zuul | 02:36 | |
| *** ChanServ sets mode: +o openstack | 02:36 | |
| *** pcaruana has joined #zuul | 05:12 | |
| openstackgerrit | neilsun proposed openstack-infra/zuul master: Using zuul string format for zuul web static_cache_expiry option https://review.openstack.org/591917 | 06:31 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: [wip] web: rewrite interface in react https://review.openstack.org/591604 | 07:21 |
| tristanC | mordred: corvus: ^ rewrote the status page in react and added stream navigation. though the ws code remains to be written. | 07:25 |
| *** lennyb has joined #zuul | 07:27 | |
| *** lennyb has quit IRC | 07:28 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: [wip] web: rewrite interface in react https://review.openstack.org/591604 | 07:28 |
| *** lennyb_ has joined #zuul | 07:34 | |
| *** lennyb_ is now known as lennyb | 07:35 | |
| *** lennyb has quit IRC | 07:35 | |
| *** electrofelix has joined #zuul | 07:40 | |
| *** lennyb_ has joined #zuul | 07:40 | |
| *** lennyb_ is now known as lennyb | 07:41 | |
| *** lennyb has quit IRC | 07:41 | |
| *** lennyb has joined #zuul | 07:44 | |
| *** lennyb has quit IRC | 08:28 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul-jobs master: npm: add CI=true environment https://review.openstack.org/591964 | 08:37 |
| *** andreaf has quit IRC | 09:16 | |
| *** mattclay_ has joined #zuul | 09:17 | |
| *** andreaf has joined #zuul | 09:17 | |
| *** mattclay has quit IRC | 09:18 | |
| *** robled has quit IRC | 09:18 | |
| *** mattclay_ is now known as mattclay | 09:18 | |
| SpamapS | man, it helps if you actually push your images into the registries you created, and not some random test string named one that you had hard coded | 10:18 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: [wip] web: rewrite interface in react https://review.openstack.org/591604 | 10:39 |
| *** panda|ruck is now known as panda|ruck|lunch | 11:25 | |
| *** panda|ruck|lunch is now known as panda|ruck | 11:53 | |
| *** rlandy has joined #zuul | 12:31 | |
| *** samccann has joined #zuul | 12:50 | |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul master: Add npm lint target back https://review.openstack.org/592022 | 13:01 |
| mordred | tristanC: cool! there ^^ is a followup patch adding a lint target back (thought it would be a good way to learn some things) | 13:01 |
| mordred | tristanC: there's still about 10 linter errors with that that don't make sense to me yet but might make sense to you | 13:02 |
| mordred | git diff | 13:02 |
| mordred | gah | 13:02 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul master: Add npm lint target back https://review.openstack.org/592022 | 13:13 |
| SpamapS | mordred: so.. now.. the container images.. | 13:14 |
| mordred | SpamapS: talk to me | 13:14 |
| SpamapS | mordred: I am running into things like, there's no /var/run/nodepool | 13:14 |
| SpamapS | Before I run off and paper over all of that.. do you have suggestions? | 13:14 |
| mordred | SpamapS: well - I think I'd been expecting that stuff like that would be attached as volumes? | 13:15 |
| SpamapS | /var/run should actually clear out when the container dies, so it should not be a volume. | 13:15 |
| SpamapS | Not even an emptyDir | 13:16 |
| mordred | what are we putting in /var/run ? | 13:16 |
| SpamapS | nodepool wants to put a pid file there | 13:16 |
| SpamapS | so I can probably find a place for it that exists | 13:16 |
| mordred | ah - yes. that pid file doesn't seem useful in a container context does it? | 13:16 |
| SpamapS | and not care | 13:16 |
| SpamapS | because I don't actually need a pid file | 13:16 |
| SpamapS | but.. papering. | 13:16 |
| mordred | SpamapS: that's part of the fork/daemonization right? would it be better to have k8s run it as a foreground process? (just thinking out loud, I don't know the answers) | 13:17 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul master: Add npm lint target back https://review.openstack.org/592022 | 13:37 |
| mordred | tristanC: ok - a few more tweaks there ^^ I fixed some of the issues, but also properly enabled the standardjs rules, so it produced a few more errors :) | 13:37 |
| openstackgerrit | Merged openstack-infra/zuul master: Remove non-working html code in div container https://review.openstack.org/591459 | 13:38 |
| sshnaidm | does anybody know when during the job job_timeout starts to count? http://logs.openstack.org/95/586595/1/check/tripleo-ci-centos-7-containers-multinode/b8a44d3/job-output.txt.gz | 13:40 |
| sshnaidm | if it from very start of zuul ececution or after some playbooks? | 13:40 |
| SpamapS | mordred: Yeah I'm wondering if we might want some standard interface for pbrx to run the entry points with some helpful hints that they should assume container. | 14:04 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul master: Add npm lint target back https://review.openstack.org/592022 | 14:04 |
| SpamapS | mordred: or have some kind of way to express final steps to prep the container | 14:05 |
| Shrews | sshnaidm: i *think* zuul.timeout starts includes both pre and run playbooks. zuul.post_timeout is only the post playbook | 14:06 |
| SpamapS | For this one, we could probably just add an os.makedirs before the pidfile? | 14:06 |
| mordred | SpamapS: I think I'd prefer the first one, cause the second one feels like it would start getting complex | 14:06 |
| sshnaidm | SpamapS, so it's from very beginning of the job, right? And is there way to find out when job started from within the job itself? | 14:07 |
| mordred | SpamapS: maybe so? could also be helpful on non-container deploys if someone has an ephemeral /var/run filesystem and the box reboots | 14:07 |
| SpamapS | sshnaidm: 301 Shrews | 14:07 |
| SpamapS | mordred: yeah, generally you aren't running nodepool-launcher as root, so that makedirs is going to fail | 14:08 |
| sshnaidm | SpamapS, oops, sorry :) | 14:08 |
| sshnaidm | Shrews, ^^ | 14:08 |
| SpamapS | sshnaidm: no worries. :) | 14:08 |
| SpamapS | it happens | 14:08 |
| sshnaidm | SpamapS, we are a lot in the "S-" club | 14:08 |
| SpamapS | mordred: this is a fairly common tactic btw.. https://github.com/kubernetes/contrib/blob/master/statefulsets/zookeeper/Dockerfile#L48-L57 | 14:09 |
| Shrews | sshnaidm: should begin at the start of pre playbook run. i'm not sure if the start time is readily available. corvus would know for sure | 14:10 |
| pabelanger | sshnaidm: Shrews: yah, job_timeout right now is shared for both pre-run / run phases. We've talked about creating a pre-run timeout before: https://git.zuul-ci.org/cgit/zuul/tree/zuul/executor/server.py#n997 | 14:11 |
| corvus | tristanC: is it not possible to use jquery with react? we want to remove it anyway, so that's fine, but i thought if jquery could be used it would make the react patch simpler since we could change the status page later. | 14:14 |
| tristanC | corvus: i think it is possible to mix both, but i would prefer we take this opportunity to clean-up the code too. | 14:15 |
| tristanC | corvus: with the ChangeQueue component, it will be possible to create a status page for a single change too | 14:16 |
| tristanC | corvus: i did keep the bootstrap classes and structure, rewritting the jquery bit in react is quite straighforward | 14:17 |
| tristanC | mordred: that's great, thanks. i'll add rebase follow-up on the linters then, and we should squash when we are satisfied | 14:18 |
| mordred | SpamapS: nod. worth pondering a bit | 14:18 |
| mordred | tristanC: yah - I think squashing is definitely easier/better than trying to keep it as a stack and rebasing | 14:19 |
| corvus | tristanC: ok; because of all the details on the status page, it might take a little longer to get this through, but if you're willing, it'll be worth it. :) | 14:19 |
| mordred | ++ | 14:19 |
| tristanC | corvus: it's mostly done, see: http://logs.openstack.org/04/591604/7/check/zuul-build-dashboard/6247836/npm/html/ | 14:20 |
| corvus | tristanC: i know, that's what i was looking at :) | 14:20 |
| corvus | tristanC: it looks good, just a few details which may or may not be easy to resolve :) | 14:20 |
| corvus | tristanC: are those glyphicons or fontawesome? | 14:23 |
| corvus | or something else? :) | 14:23 |
| pabelanger | tristanC: wow, that looks nice | 14:23 |
| tristanC | corvus: they are provided by patternfly, it mixes bootstrap and fotawesome | 14:24 |
| tristanC | pabelanger: thanks :-) | 14:24 |
| mordred | also - I'd like to note that the react patch adds javascript unittests | 14:25 |
| mordred | and a test job to run them | 14:25 |
| mordred | there is only one test - but I like that it exists and that there is a place to add more | 14:26 |
| corvus | sshnaidm, Shrews: i don't believe the start time is exposed as a var, but it would be easy to do so. | 14:26 |
| sshnaidm | corvus, where is better to do it? | 14:26 |
| mordred | tristanC: if it's easier for you to squash the lint patch sooner rather than later, please feel free - I don't feel any ownership of it, it was more just an opportunity to poke a little bit | 14:26 |
| SpamapS | mordred: to be clear.. the zk approach isn't helping either... feels like the container world just isn't in to uniformity in this area. | 14:27 |
| sshnaidm | corvus, maybe here? https://github.com/openstack-infra/zuul-jobs/blob/master/roles/emit-job-header/tasks/main.yaml#L7 | 14:27 |
| corvus | sshnaidm: if you want to add it to zuul, here would be the place: http://git.zuul-ci.org/cgit/zuul/tree/zuul/executor/server.py#n1425 | 14:28 |
| mordred | SpamapS: yah. it's an area I'd like to think about a bit as we beat our heads around this - because if the idea behind these container images we're producing is that they should be single-process style containers (as opposed to light-weight vms) - it seems like we'll want some patterns - like signaling info re: daemonization, what things should write to local disk vs. volumes, etc | 14:29 |
| fungi | anybody aware of any zuul-oriented talks which got accepted for ansiblefest? | 14:29 |
| sshnaidm | corvus, sorry, I'm not familiar with this code. When is it executed? | 14:30 |
| pabelanger | fungi: I don't think anything has officially been announced? | 14:31 |
| corvus | fungi: i think rbergeron & co were finishing up selections recently | 14:31 |
| fungi | pabelanger: yeah, i suppose at most people just know whether they've submitted any talks on the subject | 14:31 |
| fungi | thanks corvus! | 14:31 |
| corvus | sshnaidm: that's the code which creates the inventory file with the 'zuul' variables before zuul runs ansible | 14:31 |
| corvus | fungi: i know people have submitted talks on the subject! | 14:32 |
| sshnaidm | corvus, I see.. so setting timestamp there will be actually before the job itself starts | 14:32 |
| SpamapS | mordred: yeah, I wonder if we should just have a set of standard envvars, like PBRX_CONTAINER=1, that we can set in the container image, and the python can inspect to make decisions. | 14:32 |
| SpamapS | though I'm actually getting around it now with CLI args | 14:33 |
| fungi | corvus: awesome! | 14:33 |
| SpamapS | which is how zk gets around it | 14:33 |
| SpamapS | fungi: My talk was accepted! | 14:33 |
| SpamapS | I'll be sharing how we use Zuul at GoDaddy | 14:33 |
| fungi | SpamapS: at ansiblefest? neat! | 14:34 |
| corvus | mordred, SpamapS: i'm not super excited about having zuul look at PBRX env vars in the code, but it seems entirely appropriate to have a command-line or config option which causes it to run in the foreground, and perhaps some setup.cfg or whatever file that says to use that for pbrx... | 14:36 |
| corvus | tristanC, mordred: what's with the "#" in the url? | 14:37 |
| tristanC | corvus: right, so the implementation is using the HashRouter for route | 14:38 |
| tristanC | i picked this one as it will make static file simpler to use, all webclient request reach a single index.html file | 14:39 |
| tristanC | we could use the BrowserRouter to make navigation use '/' instead of '#' in the url | 14:40 |
| tristanC | but then, this need special server route to redirect the path to the root index.html | 14:40 |
| corvus | i'm a big fan of urls that are meaningful and look like urls, though the single file certainly has advantages. i'd love it if we could have both (path routing is a thing!), but if not, we'll have to think about the trade-offs. | 14:44 |
| mordred | I feel the same way - I like the current non-hash urls - but they do come with a cost, including but not limited to making it extra hard to serve multi-tenant dashboard code from swift ... | 14:45 |
| mordred | since this is using webpack under the coveres and we've solved the non-hash urls with that already I'm confident we can do it for this | 14:45 |
| mordred | but it may be worth pondering tradeoffs for sure | 14:46 |
| tristanC | it's a simple change to restore the path routing using the BrowserRouter, just need to adapt the api module to discover the base path of the api when loaded from a sub-path | 14:47 |
| tristanC | but can we at least drop the ".html" extensions and the make all the route reach a single index.html file? | 14:47 |
| corvus | tristanC: yeah, let's give that a shot and see how it looks. | 14:48 |
| corvus | (i'd love to drop the html extensions) | 14:48 |
| tristanC | mordred: i'll squash the lint and fix the last issue now | 15:40 |
| mordred | tristanC: cool | 15:43 |
| panda|ruck | pabelanger: how much you think it's difficult to make job.extra_vars: able to include variables from a file ? | 15:56 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: [wip] web: rewrite interface in react https://review.openstack.org/591604 | 15:56 |
| tristanC | corvus: mordred: last PS adds status page filters and default expand checkbox :) | 15:57 |
| panda|ruck | pabelanger: or any vars: invocation | 15:57 |
| pabelanger | panda|ruck: don't think it would be too difficult, we'd likely want to have a discussion about how to implement it first | 15:57 |
| panda|ruck | pabelanger: thanks. | 15:58 |
| corvus | tristanC, mordred: when i run 'npm run start' i get npm ERR! enoent ENOENT: no such file or directory, open '/home/corvus/git/openstack-infra/zuul/package.json' | 15:58 |
| corvus | know what i have to do to run tristanC's change? | 15:58 |
| mordred | corvus: you need to cd into the web subdir | 15:59 |
| mordred | in the patch all of the js related things are shifted into the web dir, including the package.json | 15:59 |
| corvus | mordred: oh, ok :) | 15:59 |
| corvus | that also makes yarn install do something other than exit successfully quickly | 15:59 |
| mordred | yeah. yarn install in the web dir too | 16:00 |
| tristanC | corvus: also, once you have yarn installed, you can "yarn build" or "yarn start" | 16:01 |
| corvus | tristanC: it looks like that isn't set up to use the openstack api yet, just local? | 16:02 |
| tristanC | corvus: yes, you have to use REACT_APP_ZUUL_API_ROOT="https://zuul.openstack.org/zuul/api/" environment variable | 16:02 |
| tristanC | ( https://review.openstack.org/#/c/591604/8/doc/source/developer/javascript.rst ) | 16:03 |
| tristanC | corvus: or you can wget the endpoint in the "web/public" directory and it will query the local files | 16:03 |
| mordred | we can likely add some of that back to package.json like we had before | 16:08 |
| corvus | mordred, tristanC: an interesting side effect of using glyphs instead of images for the circles is that the lines may not line up any more. it's designed as a sort of grid with the circles and lines having the same width | 16:12 |
| corvus | mordred, tristanC: maybe we can set the glyph css to a pixel value | 16:13 |
| mordred | corvus: yah - I never got the glyphs patch I tried originally to work | 16:13 |
| mordred | that's probably a good idea | 16:13 |
| corvus | tristanC: oh! the fa and pf icons are different sizes; i don't think we can mix them | 16:16 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP fix status lines https://review.openstack.org/592145 | 16:20 |
| corvus | tristanC: ^ that corrects the size and makes things line up, but we need to find 2 alternate icons | 16:20 |
| corvus | tristanC: basically, they need to render as 16x16 | 16:20 |
| corvus | the pf icons do, but not fa | 16:21 |
| corvus | but maybe there are other alternates, that's just a quick change to demonstrate it working | 16:21 |
| *** panda|ruck is now known as panda|off | 16:22 | |
| mordred | woot | 16:23 |
| mordred | corvus, tristanC: incidentally "yarn run build" or "yarn run lint" are slightly nicer output-wise than npm run build|lint ... basically because you don't get the npm exception traceback boilerplate at the end | 16:24 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: [wip] web: rewrite interface in react https://review.openstack.org/591604 | 16:25 |
| tristanC | mordred: and you can just run "yarn lint" too | 16:25 |
| tristanC | last PS adds "yarn start:openstack" and "yarn start:multi" | 16:26 |
| *** openstackstatus has joined #zuul | 16:59 | |
| *** ChanServ sets mode: +v openstackstatus | 16:59 | |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP fix status lines https://review.openstack.org/592145 | 17:09 |
| openstackgerrit | Andreas Jaeger proposed openstack-infra/zuul-website master: Link to Zuul Sphinx extension https://review.openstack.org/591282 | 17:09 |
| corvus | tristanC: there we go, i found some icons ^ that at least should fix up all the lines and icons. feel free to squash. | 17:09 |
| SpamapS | corvus: yeah I think now that I'm getting into the meat of the problem, CLI switches seem to be able to solve all my issues. | 17:10 |
| SpamapS | (regarding earlier envvar things) | 17:19 |
| SpamapS | a lot of what is getting me is that I am stealing tobiash's stuff and it has non-standard paths for stuff | 17:21 |
| tristanC | corvus: nice! i'll squash tomorrow. | 17:22 |
| SpamapS | hrm | 17:41 |
| SpamapS | having trouble getting the pidfile name changed :-P | 17:41 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Allow run to be list of playbooks https://review.openstack.org/592160 | 17:43 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Allow run to be list of playbooks https://review.openstack.org/592160 | 17:44 |
| pabelanger | mordred: corvus: tobiash: tristanC: ^propsed again for job.run to be a list of playbooks. I've included a few use cases in the commit message | 17:45 |
| pabelanger | should add reno note there too | 17:45 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul master: [WIP] Fix post pipeline change display https://review.openstack.org/592165 | 17:51 |
| mordred | tristanC: ^^ that should fix the uuid overflow in post pipeline (which makes me sad, I like the abbr idea - but it seems to be being stripped out) | 17:51 |
| mordred | pabelanger: could you say more words to me about "One example would be your job workflow would be to run multiple playbooks on the same host but use different groups in an inventory file." ? | 17:52 |
| mordred | pabelanger: (mostly my brain can't wrap itself around those specific words) | 17:53 |
| pabelanger | mordred: let me remove that, as it is not valid for me any more | 17:53 |
| mordred | ok. | 17:54 |
| mordred | I like the second one you wrote - that does seem to be in keeping with the concept "test using your production playbooks" | 17:54 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Allow run to be list of playbooks https://review.openstack.org/592160 | 17:55 |
| pabelanger | mordred: refreshed the first | 17:55 |
| pabelanger | that would be an example of tripleo does for their quickstart playbooks | 17:56 |
| corvus | mordred, pabelanger: is this better than having a job playbook which includes the setup / test playbooks? | 17:58 |
| corvus | with the existence of import_playbook it feels like borderline implementing ansible in zuul | 17:59 |
| pabelanger | corvus: that's how I done things so far today, single playbook to setup / test. However, I'm currently exploring the idea of using an external library to test, in this case testinfra. eg: https://review.openstack.org/591888/ For that, its looking more like ansible to setup then tox to test | 18:02 |
| *** sshnaidm is now known as sshnaidm|bbl | 18:05 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Allow run to be list of playbooks https://review.openstack.org/592160 | 18:10 |
| mordred | corvus: for reasons I cannot explain, I find import_playbook makes me uncomfortable | 18:13 |
| corvus | hrm. given that ansible has a construct for "run these two playbooks in series and exit immediately with failure if one of them fails", i feel like we need a compelling case for why we should reimplement that in zuul | 18:14 |
| corvus | my understanding is it would look something like this: https://etherpad.openstack.org/p/YBnjOEpHOD | 18:15 |
| pabelanger | there is also variable scoping issue with ansible and import_playbook. But running ansible-playbook twice, you don't get variables from first run in 2nd, which is how tripleo works today too | 18:15 |
| mordred | yah - I believe import_playbook is processed at compile time | 18:16 |
| corvus | now we're getting somewhere :) | 18:16 |
| corvus | so there is a substantive difference between a playbook with 2 import_playbook calls, and two separate ansible-playbook invocations? | 18:16 |
| mordred | https://twitter.com/thejimic/status/1026969663169863681 <-- jimi_c was saying that he's been resisting doing include_playbook, which would, I think, be the more similar construct | 18:16 |
| pabelanger | corvus: I believe yes, but would need to dive more into testing again. IIRC, when I last looked variables did leak across import_playbook | 18:18 |
| mordred | corvus, tristanC: hrm. I have no idea why https://review.openstack.org/#/c/592165 didn't work - I must be missing something obvious somewhere | 18:18 |
| mordred | pabelanger: like if you do a set_fact in playbook 1 I'd suppose | 18:19 |
| pabelanger | or register | 18:20 |
| pabelanger | but ya | 18:20 |
| corvus | mordred: that looks like some kind of react compile error? | 18:20 |
| corvus | that's impressive | 18:20 |
| corvus | mordred: i don't think abbr is getting eaten, it must be that changeText isn't getting set | 18:23 |
| mordred | yes. I have verified that a playbook with two import_playbooks leaks variables from the first to the second | 18:23 |
| corvus | mordred: maybe that regex test isn't working? | 18:23 |
| mordred | corvus: well, when I opened it in inspector in firefox there was no abbr ? | 18:23 |
| mordred | oh - I see what you're saying | 18:24 |
| *** electrofelix has quit IRC | 18:25 | |
| mordred | corvus: where did you see a compile error? | 18:25 |
| corvus | mordred: when i opened the test build for 592165 | 18:26 |
| corvus | i see Invariant Violation: "Minified React error #31; visit https://reactjs.org/docs/error-decoder.html?invariant=31&args[]=object%20with%20keys%20%7BchangeText%7D&args[]= for the full message or use the non-minified dev environment for full errors and additional helpful warnings. " in the console | 18:26 |
| mordred | corvus: oh neat | 18:27 |
| pabelanger | mordred: are you at a point where you can start publishing pbrx images for zuul? or is there something blocking | 18:31 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Allow run to be list of playbooks https://review.openstack.org/592160 | 18:32 |
| corvus | pabelanger, mordred: i think the standard here should be that zuul should not reipmlement things ansible already does, but allow ansible operators to do the sort of things they would do with ansible in production. it sounds with the latest info, list-of-run-playbooks meets that threshold. let's just make sure we write that up to justify it. | 18:37 |
| pabelanger | sure, let me get some coffee and I'll update the commit message | 18:38 |
| mordred | pabelanger: yes - although we need to write a pbrx push command and then we need to make a push job with the infra account credentials | 18:59 |
| mordred | pabelanger: but other than that, we're totally ready :) | 18:59 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Allow run to be list of playbooks https://review.openstack.org/592160 | 19:01 |
| pabelanger | mordred: ack! I'll look to see how loci and kolla are doing it | 19:03 |
| mordred | pabelanger: my thinking on the command, if that is a thing you want to push on - is that there is ProjectInfo class in pbrx.container_images | 19:03 |
| mordred | pabelanger: that does the work to figure out what the list of console scripts would be | 19:03 |
| mordred | pabelanger: (which in turn can provide the list of image names to push) | 19:03 |
| pabelanger | k | 19:03 |
| mordred | pabelanger: so I was thinking add a pbrx push-images command that also takes an optional --prefix argument | 19:04 |
| mordred | that would do docker push on every image name that pbrx build-images would have built | 19:05 |
| mordred | pabelanger: probably if prefix is given, we want to skip pushing the unprefixed names | 19:05 |
| mordred | pabelanger: then lifting the code from loci to set up the docker config from a zuul secret is probably a great plan | 19:06 |
| mordred | pabelanger: (also, if ou don't have time for that, I'll get around to it one of these days I promise :) ) | 19:07 |
| pabelanger | mordred: have _some_ time, so I can poke into it and see how it works | 19:11 |
| Shrews | i have some cycles (and a mostly working wrist now) I could spend on that, too if pabelanger would rather look at other things | 19:11 |
| pabelanger | Shrews: np if you want to look too | 19:12 |
| *** openstackgerrit has quit IRC | 19:19 | |
| *** sshnaidm|bbl is now known as sshnaidm | 19:22 | |
| Shrews | mordred: hrm, it's not pbrx's intent to build 2 images for each entrypoint when a prefix is supplied, is it? | 20:05 |
| Shrews | mordred: e.g., for nodepool where we have pbrx_prefix=zuul http://logs.openstack.org/12/590412/3/check/pbrx-build-container-images/12d8e91/job-output.txt.gz#_2018-08-14_15_56_03_367652 | 20:05 |
| mordred | Shrews: yah - it's probably better to only tag them with the prefix if prefix is given | 20:07 |
| *** elyezer_ has joined #zuul | 20:25 | |
| *** elyezer has quit IRC | 20:29 | |
| *** openstackgerrit has joined #zuul | 20:53 | |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul master: [WIP] Fix post pipeline change display https://review.openstack.org/592165 | 20:53 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP fix post change id link https://review.openstack.org/592197 | 20:53 |
| mordred | corvus, tristanC: ^^ ok there it is | 20:53 |
| corvus | mordred: omg lets see if we found the same thing :) | 20:53 |
| mordred | corvus: hah. jinx | 20:53 |
| corvus | er, close, but different | 20:54 |
| *** samccann has quit IRC | 20:56 | |
| corvus | mordred: yours is a more faithful port of the jquery code, i'll abandon | 20:57 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Map file comment line numbers https://review.openstack.org/590442 | 21:38 |
| openstackgerrit | Merged openstack-infra/zuul-jobs master: Add icons to index files https://review.openstack.org/591535 | 22:14 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP: Add private key storage migration https://review.openstack.org/592213 | 23:01 |
| * SpamapS fighting through making his base post job scp logs to a log storage server | 23:12 | |
| mordred | SpamapS: you've seen the logs-in-swift stuff corvus has been working on yes? | 23:49 |
| SpamapS | I have | 23:49 |
| SpamapS | it makes me hahppy | 23:49 |
| mordred | kk. just making sure | 23:49 |
| mordred | it makes me happy too | 23:49 |
| SpamapS | cause I have a big ceph cluster | 23:49 |
| SpamapS | with swift API | 23:49 |
| SpamapS | so.. +1 | 23:49 |
| mordred | although - the zuul-inline-comments makes me a bit more giddy | 23:49 |
| mordred | SpamapS: we're very close to flipping the switch for ourselves | 23:50 |
| SpamapS | what I really need next btw, is fast-forward-gate | 23:50 |
| mordred | SpamapS: I know all of those words, but I don't know what they mean in that combination | 23:50 |
| SpamapS | which basically means, for a slower-moving repo where most changes happen one at a time.. but sometimes things explode with 10... I want check->gate to be instant if it can fast forward. | 23:50 |
| mordred | oh - right - the thing darragh tweeted about | 23:51 |
| SpamapS | so that when people submit a change, it passes checks.. any jobs that are the same in check/gate, just get auto-passes if the commit fast forwards. | 23:51 |
| SpamapS | I didn't know anybody else tweeted this. | 23:51 |
| mordred | yah. electrofelix did the other day | 23:51 |
| SpamapS | But.. it just seems silly to re-run jobs on the same git state. | 23:51 |
| corvus | SpamapS: it's not silly | 23:51 |
| mordred | dependencies outside of the git code change | 23:51 |
| mordred | frequently, it turns out | 23:51 |
| SpamapS | When the job takes 2 hours and it *just* ran, it's a huge waste of my team's time. They usually just end-around and force merge. | 23:52 |
| corvus | either external dependencies, or internal (zuul-managed ones) | 23:52 |
| *** rlandy is now known as rlandy|afk | 23:52 | |
| SpamapS | Which compromises the whole gating paradigm. | 23:52 |
| corvus | SpamapS: i hear what you're saying, and i'm not saying we can't do it, but i am saying it's not silly, and it's complicated, and if we're going to do it, it's going to take some thought. | 23:53 |
| SpamapS | Totally makes sense if the check is stale or something. But I have a ton of instances where somebody makes a fix, pushes itup, checksrun, approvals happen, and then .. 2 more hours while we wait. | 23:53 |
| corvus | SpamapS: may i suggest an alternative? skip check? | 23:53 |
| mordred | yeah. on those repos skipping check and going straight to gate might be a better choice at the moment | 23:54 |
| corvus | i believe that gate requiring a check vote is an anti-pattern. don't copy that from openstack. | 23:54 |
| corvus | that's something openstack does because it operates in a specific environment where that's necessary | 23:54 |
| SpamapS | GitHub reinforces that. | 23:54 |
| SpamapS | because the thought is you need checks to be required. | 23:55 |
| SpamapS | But your point makes sense.. we could not require the openstack check for gate. | 23:55 |
| mordred | sure. but the gate votes add the checks and then do the merge | 23:55 |
| corvus | SpamapS: happily, zuul can enforce that for you instead of github | 23:55 |
| corvus | as soon as zuul splits into its own tenant, i would like to remove the clean-check requirement from zuul's own gate | 23:55 |
| clarkb | corvus: I'n not sure I would go so far as to say it is an anti pattern. Check provides extremely useful information to reviewers in many cases | 23:55 |
| corvus | then we'll be a better exemplar of the process | 23:55 |
| mordred | in fact, github's behavior + zuul without check-vote-required will make the manual "just merge it" seem more extreme | 23:56 |
| corvus | clarkb: i'm not suggesting we remove check :) | 23:56 |
| clarkb | corvus: and needing to swing back around and check if gate worked after you approve instead of just checking it upfront can be clunky | 23:56 |
| corvus | clarkb: i'm suggesting we remove the requirement. if check provides something you need (eg, review of rendered docs) by all means wait for it | 23:56 |
| clarkb | ah gotcha | 23:56 |
| mordred | like the ansible-role-puppet patch we just +A'd | 23:56 |
| SpamapS | I tend to use check passes as a gate into my review queue.. | 23:57 |
| clarkb | ya I would agree that gate requires check is a bit :( | 23:57 |
| mordred | we had to wait an extra test cycle for it even though you +A'd it 30 seconds after I pushed it up | 23:57 |
| mordred | SpamapS: me too - usually | 23:57 |
| SpamapS | But I am happy to skip that in urgent situations | 23:57 |
| mordred | but if you remove that requirement, then if a patch gets a +A before check is done, it'll go straight to gate | 23:57 |
| SpamapS | so yeah, I think not waiting for the status in gate makes sense. | 23:57 |
| mordred | ++ | 23:57 |
| SpamapS | corvus: I bow to your wisdom sir. :) | 23:57 |
| corvus | SpamapS: i accept your user requirements :) | 23:58 |
| corvus | we'll talk about the ff thing more i'm sure, i just think to do it right in zuul is going to need a bit more... computation... than it seems at first glance. and the skip-clean-check thing might be an immediate win (plus good on its own) | 23:59 |
| mordred | I had just written those words - but worse | 23:59 |
| corvus | (my guess is we'll need a real build database that we can depend on so we can query some things about history and git repo states) | 23:59 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!