| *** logan- has quit IRC | 00:12 | |
| *** logan- has joined #zuul | 00:13 | |
| *** mattw4 has quit IRC | 00:16 | |
| *** michael-beaver has quit IRC | 00:28 | |
| *** jamesmcarthur_ has quit IRC | 00:47 | |
| *** jamesmcarthur has joined #zuul | 01:01 | |
| *** jamesmcarthur has quit IRC | 01:04 | |
| *** igordc has quit IRC | 01:13 | |
| *** igordc has joined #zuul | 01:15 | |
| *** igordc has quit IRC | 01:23 | |
| *** pcaruana has quit IRC | 01:26 | |
| *** bhavikdbavishi has joined #zuul | 01:50 | |
| *** jamesmcarthur has joined #zuul | 01:56 | |
| *** jamesmcarthur has quit IRC | 01:58 | |
| *** jamesmcarthur has joined #zuul | 01:59 | |
| *** bhavikdbavishi1 has joined #zuul | 02:07 | |
| *** bhavikdbavishi has quit IRC | 02:09 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 02:09 | |
| *** jamesmcarthur has quit IRC | 02:22 | |
| *** jamesmcarthur has joined #zuul | 02:22 | |
| *** jamesmcarthur has quit IRC | 02:25 | |
| *** jamesmcarthur has joined #zuul | 02:26 | |
| *** bhavikdbavishi has quit IRC | 02:44 | |
| *** igordc has joined #zuul | 02:49 | |
| *** jamesmcarthur has quit IRC | 02:56 | |
| *** jamesmcarthur has joined #zuul | 03:12 | |
| *** pabelanger has quit IRC | 03:20 | |
| *** bhavikdbavishi has joined #zuul | 03:32 | |
| *** ianychoi_ has joined #zuul | 03:38 | |
| *** ianychoi has quit IRC | 03:42 | |
| *** pcaruana has joined #zuul | 03:44 | |
| *** tributarian has quit IRC | 03:45 | |
| *** jank has quit IRC | 03:55 | |
| *** jamesmcarthur has quit IRC | 04:00 | |
| *** jank has joined #zuul | 04:29 | |
| *** jamesmcarthur has joined #zuul | 04:30 | |
| *** jamesmcarthur has quit IRC | 04:55 | |
| *** jamesmcarthur has joined #zuul | 05:01 | |
| *** raukadah is now known as chandankumar | 05:11 | |
| *** igordc has quit IRC | 05:31 | |
| *** jamesmcarthur has quit IRC | 05:35 | |
| *** bjackman has joined #zuul | 05:49 | |
| *** bjackman has quit IRC | 06:03 | |
| *** jamesmcarthur has joined #zuul | 06:03 | |
| *** bjackman has joined #zuul | 06:19 | |
| *** jamesmcarthur has quit IRC | 06:37 | |
| *** themroc has joined #zuul | 07:05 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Use wait for empty update queue before accepting merges https://review.opendev.org/675039 | 07:10 |
|---|---|---|
| tobiash | corvus, mordred: there is another case of a missing TextWorker override ^ | 07:12 |
| *** sshnaidm|afk is now known as sshnaidm|bbl | 07:17 | |
| tobiash | mordred: are you in europe or jetlagged? | 07:19 |
| mordred | neither - just awake for no reason :) | 07:20 |
| tobiash | ;) | 07:20 |
| *** tosky has joined #zuul | 07:22 | |
| * SpamapS *feels* that | 07:23 | |
| mordred | SpamapS: otoh - it's a good time to get some things done :) | 07:24 |
| SpamapS | yerp | 07:28 |
| *** jamesmcarthur has joined #zuul | 07:35 | |
| *** spsurya has joined #zuul | 07:50 | |
| openstackgerrit | Merged zuul/zuul master: Update ara to >=0.16.5 to support ansible 2.8 https://review.opendev.org/674958 | 08:01 |
| *** jangutter has joined #zuul | 08:05 | |
| *** threestrands has quit IRC | 08:06 | |
| *** jamesmcarthur has quit IRC | 08:08 | |
| *** panda has quit IRC | 08:22 | |
| *** panda has joined #zuul | 08:23 | |
| openstackgerrit | Mark Meyer proposed zuul/zuul master: Rework a cache invalidation issue https://review.opendev.org/674425 | 08:38 |
| *** jamesmcarthur has joined #zuul | 08:42 | |
| *** jank has quit IRC | 09:19 | |
| openstackgerrit | Fabien Boucher proposed zuul/zuul master: Add reference pipelines file for Gerrit driver https://review.opendev.org/672683 | 09:20 |
| openstackgerrit | Fabien Boucher proposed zuul/zuul master: A reporter for Elasticsearch https://review.opendev.org/644927 | 09:27 |
| ofosos | I've some problems with the container builds. Apparently the containers is builds do not include all the changes I made locally. Is that possible? Do they do some git magic? | 09:39 |
| *** bhavikdbavishi has quit IRC | 09:39 | |
| *** sshnaidm|bbl is now known as sshnaidm | 09:42 | |
| *** jamesmcarthur has quit IRC | 09:57 | |
| *** jank has joined #zuul | 10:17 | |
| ofosos | All good. | 10:27 |
| openstackgerrit | Sagi Shnaidman proposed zuul/zuul-jobs master: Don't install centos repos on RHEL https://review.opendev.org/674572 | 10:36 |
| *** rfolco has joined #zuul | 10:54 | |
| *** rfolco|ruck has quit IRC | 10:54 | |
| *** jangutter_ has joined #zuul | 10:55 | |
| *** jangutter has quit IRC | 10:58 | |
| *** jangutter_ has quit IRC | 11:08 | |
| *** jangutter has joined #zuul | 11:12 | |
| *** jangutter has quit IRC | 11:14 | |
| *** jangutter has joined #zuul | 11:38 | |
| *** jamesmcarthur has joined #zuul | 11:38 | |
| flaper87 | does OpenStack's Zuul runs master? Or is it pinned to a specific commit/version? | 11:40 |
| *** jangutter has quit IRC | 11:40 | |
| flaper87 | s/runs/run/ | 11:40 |
| AJaeger | AFAIK, it runs master at some time ;) We need to restart to go there, so every so often we restart with current master... | 11:43 |
| *** jangutter has joined #zuul | 11:43 | |
| flaper87 | thanks, that's what I thought too. Wanted to confirm | 11:43 |
| *** bjackman has quit IRC | 11:45 | |
| *** jangutter has quit IRC | 11:45 | |
| *** bhavikdbavishi has joined #zuul | 11:58 | |
| *** jangutter has joined #zuul | 11:59 | |
| *** bhavikdbavishi1 has joined #zuul | 12:13 | |
| *** rlandy has joined #zuul | 12:13 | |
| *** bhavikdbavishi has quit IRC | 12:14 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 12:14 | |
| *** wxy-xiyuan has quit IRC | 12:15 | |
| *** jamesmcarthur has quit IRC | 12:16 | |
| *** rfolco has quit IRC | 12:28 | |
| *** michael-beaver has joined #zuul | 12:35 | |
| *** zbr has quit IRC | 12:46 | |
| *** jamesmcarthur has joined #zuul | 12:49 | |
| *** pabelanger has joined #zuul | 12:57 | |
| *** zbr has joined #zuul | 13:02 | |
| *** jeliu_ has joined #zuul | 13:12 | |
| *** jamesmcarthur has quit IRC | 13:13 | |
| *** jamesmcarthur has joined #zuul | 13:14 | |
| fungi | flaper87: AJaeger: yeah, we continuously *install* master branch tip onto our zuul servers, but we don't end up running every single commit since we can sometimes go weeks between restarts | 13:15 |
| fungi | however, we *do* as a rule restart everything and run for a couple days on what you might consider a "release candidate" commit before tagging a release | 13:16 |
| fungi | also, not to be pedantic, it's no longer openstack's zuul (at least in name). it's now the opendev zuul and services quite a number of non-openstack projects (always has really, though openstack's workload still consumes a vast majority of the resources) | 13:17 |
| openstackgerrit | Jeff Liu proposed zuul/zuul-operator master: WIP: Add zuul-operator-functional-openshift job https://review.opendev.org/674355 | 13:38 |
| *** jamesmcarthur has quit IRC | 13:44 | |
| *** jamesmcarthur has joined #zuul | 13:46 | |
| *** electrofelix has quit IRC | 13:50 | |
| *** jamesmcarthur has quit IRC | 13:51 | |
| *** jamesmcarthur has joined #zuul | 13:53 | |
| *** jank has quit IRC | 14:19 | |
| tristanC | Hi, does config projects job running in post-review pipeline are executed speculatively? Or does the post-review only affects secrets and project ssh key usage? | 14:20 |
| pabelanger | not speculatively, need to merge the code first | 14:22 |
| tristanC | pabelanger: alright, thanks! | 14:23 |
| corvus | tristanC, mordred, anyone else: i need some feedback from you. can you please visit all of the preview sites listed in this ethercalc and leave review comments on the associated changes letting me know what you like / don't like about them? https://ethercalc.openstack.org/yz239hzwvphk | 14:38 |
| tristanC | corvus: i'll review them now | 14:42 |
| *** mattw4 has joined #zuul | 14:46 | |
| *** altlogbot_3 has quit IRC | 14:46 | |
| fungi | tristanC: pabelanger: post-review is orthogonal to speculative execution, i'm pretty sure. things can be speculatively executed in opendev's "gate" pipeline (assuming the right conditions are met) even though it's marked "post-review: True" https://opendev.org/openstack/project-config/src/branch/master/zuul.d/pipelines.yaml#L71 | 14:46 |
| fungi | tristanC: oh, though you said "config projects" and yes, those won't have speculative execution regardless of whether the pipeline is set to post-review (still the two are unrelated as far as i know) | 14:47 |
| mordred | corvus: looking | 14:47 |
| *** altlogbot_2 has joined #zuul | 14:48 | |
| *** jamesmcarthur has quit IRC | 14:51 | |
| mordred | corvus: I'm going through and leaving review comments this pass in isolation - like, not considering the other patches - I'm not sure if that'll be relevant | 14:53 |
| *** ianychoi_ is now known as ianychoi | 14:54 | |
| *** mattw4 has quit IRC | 14:55 | |
| corvus | mordred: yeah, some of them can work in isolation, others work better or require other patches... we'll figure it out :) | 14:57 |
| corvus | also, i can't access the ethercalc now, can anyone else? | 14:57 |
| mordred | nope. | 15:00 |
| *** mattw4 has joined #zuul | 15:02 | |
| *** mattw4 has quit IRC | 15:08 | |
| corvus | it's back | 15:08 |
| *** mattw4 has joined #zuul | 15:09 | |
| *** rlandy is now known as rlandy|rover | 15:13 | |
| *** chandankumar is now known as raukadah | 15:13 | |
| *** themroc has quit IRC | 15:16 | |
| *** jamesmcarthur has joined #zuul | 15:18 | |
| mordred | corvus: on https://review.opendev.org/#/c/674951/ I'm seeing https://i.imgur.com/hwOYTdT.png | 15:25 |
| sshnaidm | pabelanger, can you take a look please? https://review.opendev.org/#/c/674572 | 15:25 |
| mordred | corvus: also - tasks that don't have any stdout still have an expansion arrow, but expanding it just renders an empty box | 15:27 |
| mordred | (that's across teh board) | 15:27 |
| corvus | mordred: re the first thing: yeah, that's just a task without a name | 15:28 |
| corvus | mordred: we need to have that fall back to something, like "command". but that's not related to that change, that's true for all of these dashboards | 15:28 |
| AJaeger | sshnaidm: where was 674572 tested inside OpenDev, e.g. with OpenStack jobs? | 15:29 |
| corvus | mordred: and yes, the expansion arrow thing is annoying. it's a defect in patternfly react. we'll have to get an upstream fix in, but we may be able to work around it locally with a whole bunch of copypasta. | 15:29 |
| mordred | corvus: oh joy | 15:29 |
| sshnaidm | AJaeger, it's used in every tripleo job | 15:29 |
| AJaeger | could you make a dummy tripleo change to test it, please? | 15:30 |
| sshnaidm | AJaeger, sure | 15:31 |
| sshnaidm | AJaeger, https://review.opendev.org/#/c/674885/ | 15:31 |
| AJaeger | sshnaidm: thanks, let's add that as comment to the review as well... | 15:32 |
| sshnaidm | AJaeger, done | 15:33 |
| mordred | corvus: on that one that I commented about it not working ... I see what my issue was | 15:44 |
| mordred | corvus: I picked a multinode job, but it was a legacy grenade job | 15:44 |
| corvus | yeah, those have terrible descriptions :) | 15:45 |
| corvus | mordred, tristanC: the gist i'm getting is that all those changes look good, and there are some suggestions to further tweak them (eg, visually split plays without nesting, add icon for hostname) | 15:45 |
| mordred | corvus: so the tasks I was looking at just didn't have descriptions, and that combined with teh "if there's no stdout the task still expands" | 15:45 |
| corvus | and oh yeah, i can try to work around the expansion issue | 15:45 |
| corvus | (it's going to be a lot of boilerplate code copied from patternfly react though :( ) | 15:45 |
| mordred | it looked like "description -> empty" "(no description) -> output" "(no description) -> output" | 15:46 |
| mordred | corvus: that's gross :( | 15:46 |
| corvus | yeah. i'll make sure to open a PR. | 15:46 |
| mordred | corvus: is it a hard upstream fix? | 15:46 |
| mordred | cool | 15:46 |
| corvus | nope, easy | 15:46 |
| corvus | they're just missing an empty div. | 15:46 |
| mordred | maybe we can convince them to land a pr and cut a bugfix release for us :) | 15:46 |
| corvus | now, they *might* go and say that merging it would change everyone's rendering and decline it. but that's the wrong approach -- their code doesn't match the patternfly docs. | 15:47 |
| corvus | hopefully they'll see it our way :) | 15:47 |
| mordred | corvus: anywho - check out neutron-grenade-multinode for a patholigical case | 15:47 |
| mordred | corvus: I think, especially with the "move the hostname" patch, we might want to put something in description if the playbook doesn't have something there | 15:48 |
| corvus | mordred, tristanC: should i just squash all of those changes into one for ease of merging? or do we want to just approve them as they are? either way, i'll work on implementing the suggestions as followups. | 15:48 |
| corvus | mordred: absolutely. we can at least put the task action there. | 15:49 |
| mordred | corvus: yeah - I think the whole stack is good | 15:49 |
| mordred | corvus: actually - the moved hostname patch makes missing description look less bad :) | 15:50 |
| mordred | corvus: oh - I just noticed ... | 15:50 |
| mordred | the i in a circle next to trusted - my expectation on seeing it would be that I could click the i to get further information | 15:51 |
| mordred | (that might just be how my eyes see the ui) - but maybe a followup could make a click on the i pop up a description of what trusted means? | 15:51 |
| *** jamesmcarthur has quit IRC | 15:52 | |
| corvus | mordred: on the summary tab, we have task counts in a similar place and they aren't clickable | 15:52 |
| corvus | though i agree that once we make the "SKIPPED" etc labels clickable, it gets less clear | 15:52 |
| mordred | corvus: so - on the summary page - mouseovers tell me what the i check and x mean | 15:53 |
| mordred | (although now that I'm looking at *that* - i == ok tasks and check == changed tasks - but on the console page changed is yellow and ok is green) | 15:54 |
| corvus | oh, another difference is that the entire row is clickable because of the expansion, so you don't have mouse-cursor feedback telling you that the (i) isn't clickable | 15:54 |
| mordred | yeah | 15:54 |
| corvus | tricky one this. | 15:54 |
| mordred | I think mouseover help would be fine (right now i on console just says "trusted") | 15:54 |
| mordred | yeah | 15:54 |
| mordred | ui is fun | 15:54 |
| corvus | okay, i'm going to squash everything and start on updates | 15:55 |
| mordred | cool | 15:55 |
| openstackgerrit | James E. Blair proposed zuul/zuul master: Render console in js https://review.opendev.org/674368 | 15:55 |
| mordred | corvus: I think this is looking REALLY good and super helpful | 15:55 |
| corvus | \o/ thanks for your feedback -- i need extra eyes :) | 15:56 |
| corvus | i think we can go ahead and merge that at this point if we wanted ^ | 15:56 |
| corvus | mordred: i don't suppose we have npm sibling support? :) | 15:58 |
| corvus | was wondering if i could do the pf-react PR and depends-on | 15:59 |
| *** igordc has joined #zuul | 16:07 | |
| mordred | corvus: we do not have npm sibling support - but I thnik we should :) | 16:10 |
| mordred | corvus: have you done yarn overrides of dependenices locally yet? I think I read about it a little while ago | 16:10 |
| mordred | corvus: https://yarnpkg.com/en/docs/cli/link | 16:12 |
| corvus | mordred: no. thanks for that | 16:13 |
| mordred | corvus: so - it looks like it should be fairly easy-ish to make an npm-siblings | 16:13 |
| mordred | corvus: however, it might be a little weird and/or different for zuul since the npm dir is down a dir | 16:13 |
| mordred | corvus: so maybe we just do a local to zuul version to start - with an explicit list of deps | 16:14 |
| mordred | and figure out generalized version later | 16:14 |
| mordred | corvus: also - fwiw - we could do siblings and depends-on to test that it works - and if we wanted to run with it while waiting, we could just update our packages.json to point to your gh repo / branch with the patch in it until they make a release | 16:21 |
| *** jamesmcarthur has joined #zuul | 16:22 | |
| tosky | it looks like https://review.opendev.org/#/c/674334/ should be working fine now (finally), sorry for the messy 1st try | 16:28 |
| openstackgerrit | James E. Blair proposed zuul/zuul master: Usability tweaks for the build page console https://review.opendev.org/675147 | 16:31 |
| corvus | mordred, tristanC: ^ that's everything except the expansion fix | 16:31 |
| *** igordc has quit IRC | 16:51 | |
| *** igordc has joined #zuul | 17:10 | |
| *** jamesmcarthur has quit IRC | 17:12 | |
| *** altlogbot_2 has quit IRC | 17:17 | |
| corvus | mordred: on second inspection, the patternfly documentation is a bit ambiguous about whether the heading should be aligned for rows with and without expansion carets. on the plus side, there are two examples, one with a checkbox, and one with a caret and checkbox, and the headers align. on the other hand, the examples don't explicitly include the div for the caret if it isn't there, and they also say don't | 17:21 |
| corvus | mix-and-match row styles. at any rate, there isn't a clear-cut thing i can point to to make the argument for the change, so i'm not sure it's worth it. however, i have managed to accomplish the same result with just a slight css tweak, so i think that'll do for now. | 17:21 |
| corvus | https://www.patternfly.org/v3/pattern-library/content-views/list-view/#code (is what i was looking at) | 17:22 |
| corvus | almost ready to push that up | 17:22 |
| *** altlogbot_3 has joined #zuul | 17:24 | |
| *** jamesmcarthur has joined #zuul | 17:28 | |
| *** altlogbot_3 has quit IRC | 17:30 | |
| corvus | mordred: oh, i found i can change the cursor for the (i) icon | 17:32 |
| *** jamesmcarthur has quit IRC | 17:32 | |
| mordred | corvus: neat! | 17:33 |
| *** jamesmcarthur has joined #zuul | 17:34 | |
| AJaeger | here're two changes for zuul-jobs that could introduce problems - https://review.opendev.org/674572 and https://review.opendev.org/#/c/674334/ . I'd like to have a second/third pair of eyes on them if anybody has time... | 17:35 |
| *** altlogbot_1 has joined #zuul | 17:36 | |
| openstackgerrit | James E. Blair proposed zuul/zuul master: Don't always show expansion option on build console https://review.opendev.org/675163 | 17:36 |
| corvus | mordred, tristanC: ^ i think that's everything from the earlier round of comments | 17:37 |
| * corvus afks for a bit | 17:37 | |
| *** mattw4 has quit IRC | 17:38 | |
| *** jamesmcarthur has quit IRC | 17:39 | |
| *** jamesmcarthur has joined #zuul | 17:49 | |
| *** jamesmcarthur has quit IRC | 18:00 | |
| *** jamesmcarthur has joined #zuul | 18:01 | |
| mnaser | pabelanger: do you have any jobs that you can share that leverage ansible-galaxy-import ? | 18:01 |
| *** sgw has quit IRC | 18:03 | |
| pabelanger | mnaser: https://github.com/ansible-network/zuul-config/blob/32d9815273e586147fefe560985a6b1a1a9425b4/playbooks/publish/galaxy.yaml | 18:04 |
| pabelanger | https://github.com/ansible-network/zuul-config/blob/32d9815273e586147fefe560985a6b1a1a9425b4/zuul.d/jobs.yaml#L72 | 18:04 |
| pabelanger | but with next version of ansible, that changes I _think_ | 18:05 |
| pabelanger | I'm not sure what is going to happen with import command | 18:05 |
| pabelanger | since galaxy will be able to upload a tarball for collections | 18:05 |
| mnaser | ah i see | 18:06 |
| mnaser | so maybe ill hold off that for now i guess | 18:06 |
| pabelanger | collections is the new hotness, but something still being worked out | 18:06 |
| mordred | pabelanger: when you say "be able to upload a tarball" - does that also mean there's support for not-github now then? | 18:07 |
| pabelanger | mordred: yes, but upload of tarballs are for collections only. I need to check how a role would work | 18:08 |
| pabelanger | I am guessing that is still import via github | 18:08 |
| pabelanger | yah, ansible-galaxy role import is still a thing, which is github online | 18:09 |
| pabelanger | only* | 18:09 |
| mordred | pabelanger: oh - that's fine - if collections support not-github, then I'll be much more interested in publishing collections than roles :) | 18:11 |
| pabelanger | yah, I suspect you'll want to be involved with openstack collection :) | 18:12 |
| fungi | now if they can just work out how to support hyphens | 18:15 |
| pabelanger | yah, I don't expect that to be fixed anytime soon, TBH | 18:16 |
| pabelanger | which means I get to do 'fun' things like: https://github.com/ansible/ansible-zuul-jobs/pull/166/files#diff-51cb23550bb81f6d4618f9f610a78ec0R20 | 18:16 |
| *** bhavikdbavishi has quit IRC | 18:16 | |
| corvus | pabelanger: are you using the upload-logs-swift role directly from zuul-jobs master? | 18:19 |
| pabelanger | corvus: yes, I believe so | 18:19 |
| pabelanger | yah, no local fork | 18:20 |
| *** sgw has joined #zuul | 18:21 | |
| corvus | pabelanger: what is logs.zuul.ansible.com ? | 18:22 |
| *** spsurya has quit IRC | 18:22 | |
| pabelanger | corvus: nginx server to rewrite swift url | 18:22 |
| corvus | pabelanger: why rewrite the swift url? | 18:22 |
| corvus | (for comparison, here's a build in opendev with a bare swift url for its log url: http://zuul.opendev.org/t/zuul/build/1e202ae3ea924496941dfd56ce7d372a ) | 18:24 |
| pabelanger | humans wanted a more friendly url then https://object-storage-ca-ymq-1.vexxhost.net/v1/a0b4156a37f9453eb4ec7db5422272df/logs/ | 18:24 |
| pabelanger | it basically does proxy pass with SSL | 18:24 |
| pabelanger | running on same system as zuul-web | 18:24 |
| corvus | pabelanger: okay. i'm probably not going to understand why people think https://logs.zuul.ansible.com/86/60086/5555cc8f2d1b7919ef061b797256767c8109c061/third-party-check/ansible-test-network-integration-vyos-python37/81e293a/ is better than | 18:26 |
| corvus | https://object-storage-ca-ymq-1.vexxhost.net/v1/a0b4156a37f9453eb4ec7db5422272df/logs/86/60086/5555cc8f2d1b7919ef061b797256767c8109c061/third-party-check/ansible-test-network-integration-vyos-python37/81e293a/ but maybe it's not important that i do. :) | 18:26 |
| corvus | pabelanger: i'd like for the ansible zuul instance to take advantage of all the new shiny log stuff we've been working on | 18:26 |
| corvus | pabelanger: there's an easy first step, and that's to add a cors header to that nginx config. that should be all that's needed | 18:27 |
| pabelanger | yah, i don't have a strong opinion myself. I can bring it up at next team meeting and see what other folks want to do | 18:27 |
| corvus | pabelanger: once that's in place, you'll be ready to switch to having zuul report the build page url instead of the logs url | 18:27 |
| pabelanger | it was just confusing to explain what vexxhost was to people when talking about testing | 18:27 |
| pabelanger | k | 18:27 |
| corvus | pabelanger: then once you do that, you might then think about removing the nginx proxy -- the swift urls won't show up as prominently (they will only be visible if someone navigates through a couple of layers to go directly to a hosted logfile) | 18:28 |
| pabelanger | ack | 18:29 |
| pabelanger | also, does https://zuul.openstack.org/stream/53c1b50d74d546eeadc44f53aef9b06a?logfile=console.log work for anybody else? | 18:29 |
| pabelanger | I don't seem to be seeing logs stream properly | 18:29 |
| corvus | pabelanger: i wouldn't worry about changing the nginx thing until then -- it's not worth it until then, and once you get to the end state, it'd probably be an easy sell | 18:29 |
| corvus | pabelanger: we're having a bunch of oom-kills of the streaming daemon because of the executor distribution bug | 18:29 |
| corvus | the bug is fixed but we haven't restarted executors with it yet | 18:29 |
| pabelanger | corvus: yah, I do for see a big issue with the change. And happy to do it | 18:30 |
| pabelanger | ah, missed that. Thanks | 18:30 |
| corvus | pabelanger: i assume you meant you "don't" foresee a big issue? :) | 18:30 |
| corvus | pabelanger: where's the config for the nginx server? | 18:30 |
| pabelanger | sorry, don't yes | 18:30 |
| pabelanger | corvus: it should be https://github.com/ansible-network/windmill-config but doesn't look like it is properly manged via gitops :( | 18:31 |
| pabelanger | so maybe a good reason to remove it | 18:31 |
| pabelanger | http://paste.openstack.org/show/755632/ | 18:31 |
| pabelanger | that is the contents | 18:31 |
| corvus | pabelanger: i'll send you a PR after i get back from lunch... and you can deploy that however you want :) | 18:32 |
| pabelanger | ack | 18:32 |
| corvus | pabelanger: it should be a one liner, so i'd still say the easiest course is to keep nginx around at least until after the switch to reporting the build page | 18:32 |
| corvus | then getting rid of it should be easier | 18:33 |
| *** jeliu_ has quit IRC | 18:33 | |
| corvus | pabelanger: getting rid of logs.opendev.org is a goal for us, so the process matches up pretty well :) | 18:33 |
| pabelanger | yah, removing is likely best now | 18:33 |
| pabelanger | it* | 18:33 |
| *** mattw4 has joined #zuul | 18:33 | |
| pabelanger | corvus: in fact, we can test it now in https://github.com/ansible/project-config/blob/master/playbooks/base-minimal-test/post-logs.yaml by removing logs.o.o from the base job | 18:35 |
| corvus | pabelanger: i think the swift container is missing some cors setting for that to work | 18:37 |
| corvus | so it's better to keep nginx until we fix that | 18:37 |
| pabelanger | ok | 18:37 |
| corvus | gotta grab lunch now, biab. | 18:37 |
| *** mattw4 has quit IRC | 18:44 | |
| *** mattw4 has joined #zuul | 18:44 | |
| *** jamesmcarthur has quit IRC | 19:03 | |
| pabelanger | tobiash: first pass on https://review.opendev.org/674854/ is green, which is good. that is ansible 2.9 support. Still need to dig into test results | 19:05 |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Be consistent about spaces before and after vars https://review.opendev.org/667698 | 19:16 |
| *** jamesmcarthur has joined #zuul | 19:16 | |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Be consistent about spaces before and after vars https://review.opendev.org/667698 | 19:17 |
| tobiash | pabelanger: cool :) | 19:22 |
| tobiash | corvus: we also have a proxy in front of our swift logs because of auth reasons. I guess this won't ne a problem? Adding any headers there is no problem for me. | 19:24 |
| pabelanger | Oh, that was the other reason for us too. | 19:25 |
| pabelanger | there was talk of doing private data on public zuul, and needed a way to control auth | 19:25 |
| pabelanger | for things like logs | 19:25 |
| pabelanger | and status UI | 19:25 |
| pabelanger | but, that seems to have staled out for now | 19:25 |
| *** tributarian has joined #zuul | 19:27 | |
| corvus | tobiash: yeah, i expect it should be fine. we'll have a demo soon | 19:30 |
| tobiash | cool | 19:31 |
| corvus | also i plan on sending an email to the list about this, it's just i'm not sure what to write yet -- i'd like to find out exactly what we would need to do to fix the swift container ansible's zuul is using | 19:31 |
| *** jamesmcarthur has quit IRC | 19:35 | |
| *** jamesmcarthur has joined #zuul | 19:36 | |
| corvus | pabelanger, tobiash: https://github.com/ansible-network/windmill-config/pull/487 should do it i think | 19:36 |
| corvus | the easiest way to tell if it's working is to visit a build page like this: https://dashboard.zuul.ansible.com/t/ansible/build/81e293adb9294c35a8d17a22e135a5a5 | 19:37 |
| corvus | if you see the host summary information at the bottom, it's working. see http://zuul.opendev.org/t/zuul/build/1e202ae3ea924496941dfd56ce7d372a for comparison | 19:37 |
| pabelanger | Oh, it is in git, Yay | 19:38 |
| pabelanger | approved, should know more in 10mins | 19:40 |
| corvus | pabelanger: what's the base job in that system? | 19:41 |
| corvus | i found ansible/project-config, but i only see base-minimal in there, and i don't see the option to set that as the default base job for the tenant | 19:42 |
| pabelanger | yah, base-minimal | 19:42 |
| pabelanger | let me check | 19:42 |
| pabelanger | I don't think we changed the default base job, we just parent base to base-minimal | 19:43 |
| pabelanger | and it seems to work | 19:43 |
| corvus | where's base? | 19:43 |
| pabelanger | https://github.com/ansible/ansible-zuul-jobs | 19:43 |
| corvus | pabelanger: oh, that's an untrusted repo? | 19:44 |
| pabelanger | yah | 19:44 |
| pabelanger | base is untrusted for us, and base-minimal trusted | 19:45 |
| corvus | ok | 19:46 |
| corvus | the 'configure-mirrors-fork' role makes me think we should make sure you review the new mirror replacement stuff: https://zuul-ci.org/docs/zuul-jobs/mirror.html | 19:47 |
| pabelanger | yah, there has been some things we forked for speed mostly. We also don't have a good story for mirror infra in zuul.ansible.com | 19:48 |
| fungi | it's possible the auth problem could be worked around by the fact that the build ids should be unguessable? so your only real risk of disclosure (as brute-forcing a uuid isn't viable) is someone leaking the url somewhere outside of the system | 19:49 |
| fungi | which i suppose could still be a risk | 19:49 |
| corvus | fungi: yeah, though the log urls are predictable (but they don't need to be) | 19:50 |
| corvus | we just liked them that way | 19:50 |
| fungi | oh, right, and we only embed an abbreviated build id don't we | 19:50 |
| corvus | we could totally add a flag to upload-logs-swift that makes the urls unguessable | 19:50 |
| fungi | the abbreviated build id would be fairly trivial to brute-force | 19:51 |
| corvus | pabelanger: https://github.com/ansible/project-config/pull/171 | 19:54 |
| corvus | pabelanger: and https://github.com/ansible/project-config/pull/172 | 19:55 |
| corvus | pabelanger: not sure the best way to test 171, i'll leave that up to you :) | 19:56 |
| corvus | also, i really have no idea if that's the right way to create a patch series in github | 19:56 |
| *** jeliu_ has joined #zuul | 19:56 | |
| pabelanger | yah, can't really. they expect you to push it up into single PR | 19:56 |
| pabelanger | we'll have to rebase 172 once 171 merges | 19:57 |
| mordred | corvus: you can depends-on between the two to create a patch series | 19:57 |
| mordred | I mean - from a zuul pov at least | 19:57 |
| pabelanger | Yah, about the only way to really do it | 19:58 |
| mordred | github won't understand it - but at least it won't show the diff from 171 in 172 | 19:58 |
| corvus | pabelanger: huh, really? 172 won't merge after 171? | 19:58 |
| pabelanger | corvus: no, will be merge conflict as they have same commit | 19:58 |
| corvus | it should just be 2 fast-forwards | 19:58 |
| corvus | wow | 19:58 |
| corvus | that is so not how git works | 19:58 |
| pabelanger | I think, it does | 19:58 |
| pabelanger | lets find out | 19:58 |
| corvus | yeah, good experiment | 19:58 |
| mordred | corvus: they tend to use the rebase/squash merge method in ansible | 19:59 |
| corvus | (since it's a low-traffic repo, we shouldn't have any merges) | 19:59 |
| mordred | or the chery-pick or whatever it's called | 19:59 |
| mordred | I don't know if that's the case in that repo - I know it is in ansible/ansible | 19:59 |
| pabelanger | it will create a merge commit I think | 19:59 |
| corvus | mordred: oh that would botch it for sure | 19:59 |
| pabelanger | but lets see | 19:59 |
| mordred | but yes - good experiement! | 19:59 |
| corvus | mordred: cause yeah, that would make a new commit with identical content | 19:59 |
| mordred | yah | 19:59 |
| mordred | which also a zuul-only depends-on patch series would handle | 20:00 |
| corvus | mordred, tristanC: can you take a look at the site preview build in https://review.opendev.org/675163 and vote on it and also the 2 parent changes? | 20:02 |
| corvus | there's a tiny lint issue on that one which i'll fix now | 20:03 |
| openstackgerrit | James E. Blair proposed zuul/zuul master: Don't always show expansion option on build console https://review.opendev.org/675163 | 20:03 |
| corvus | that was an extra semicolon, that's not going to change anything | 20:04 |
| corvus | mordred, tristanC, clarkb: i'm really digging the console thing -- you have all come up with some great ideas :) | 20:05 |
| mordred | \o/ | 20:05 |
| corvus | now i'm going to try fixing some of the deep-link issues clarkb pointed out yesterday | 20:05 |
| mordred | corvus: if all I have to do is say random things in irc and you write a bunch of javascript, that works great for me | 20:05 |
| corvus | mordred: that's cute that you think that's all you have to do :) | 20:06 |
| * mordred batts eyelashes | 20:06 | |
| pabelanger | corvus: https://dashboard.zuul.ansible.com/t/ansible/build/81e293adb9294c35a8d17a22e135a5a5 looks to be correct now | 20:07 |
| pabelanger | thanks! | 20:07 |
| corvus | pabelanger: \o/ | 20:08 |
| corvus | pabelanger: the js web app is running off of releases too, right? (ie, it's not from master?) | 20:08 |
| pabelanger | yah, released version | 20:09 |
| fungi | not to distract, but there's an osf newsletter going out this week... is there anything we think might be worth mentioning (aside from the impending ansiblefest)? i'm guessing mentioning the new dashboard work might still be premature for that audience? | 20:09 |
| corvus | fungi: yeah, let's flag that for the next one; we'll have nice things to show off | 20:09 |
| fungi | i figured, but wanted to make sure. thanks! | 20:10 |
| fungi | in a couple more weeks there's a good chance it'll be in a release i guess | 20:10 |
| fungi | and that makes it a lot easier to talk about | 20:10 |
| mordred | corvus: I'm not seeing console in the latest dashboard build | 20:10 |
| corvus | pabelanger: after getting the manifest role in place, the next step for ansible's zuul will be to wait for a release with the logs and console tab. i think we can do that next week. | 20:10 |
| mordred | corvus: nevermind please | 20:11 |
| corvus | pabelanger: then we can verify that all works, then you'll have the option to switch the reporting url to the build page | 20:11 |
| pabelanger | corvus: we could do pre-release from pypi too, and that will pick up everything from wheel | 20:11 |
| corvus | (which is a change i still have to right) | 20:11 |
| corvus | er, to write | 20:11 |
| corvus | pabelanger: yeah, though we're way past-due for a release and i think making one next week after we land the (yet-to-be-written) change to add the feature flag for the reporting url is a really good place to do one. plus what fungi said. | 20:12 |
| pabelanger | wfm! | 20:12 |
| mordred | corvus: oh wow. something went poorly | 20:12 |
| corvus | (i *think* we've merged the initial batch of web auth stuff; i didn't want to have that span releases) | 20:13 |
| fungi | mordred: i hope that was an actual error message. if so, i love those | 20:13 |
| corvus | fungi: actually it's "Something went wrong" i think | 20:13 |
| mordred | https://i.imgur.com/ssP2kiF.png | 20:13 |
| mordred | corvus: ^^ | 20:13 |
| corvus | but we can totally change it to 'poorly' | 20:13 |
| fungi | heh | 20:13 |
| mordred | those results numbers are really big :) | 20:13 |
| corvus | mordred: yes, though to be fair, that looks a lot better when there's actually data | 20:13 |
| corvus | that'll be, it looped over 3 things, none of which actually output anything interesting | 20:14 |
| mordred | corvus: joy :) | 20:14 |
| corvus | so 2 questions: | 20:14 |
| fungi | oh, i see now, when you said the numbers are really big, you meant visually, not numerically | 20:14 |
| corvus | 1) should we hide the result item header when there isn't anything to display under it? | 20:14 |
| mordred | corvus: the rest of things look *great* | 20:15 |
| corvus | 2) should we adjust the header sizes? (make those smaller, make the other ones bigger?) but we should evaluate this question on a task with actual data and item names. better to find something in the first pre-playbook, that should look right. | 20:15 |
| fungi | ("hide" could also just be "de-emphasize somehow" i guess, like pale color, shade background, shrink...?) | 20:16 |
| mordred | yeah. that was again from a legacy job - so god only knows what's going on | 20:16 |
| mordred | corvus: stack is +2 from me | 20:16 |
| corvus | mordred: first pre-playbook "Print node information" task is a good one | 20:16 |
| corvus | i think i lean toward "yes" for question 1 -- so in mordred's screenshot, we would just not have "results" or anything under it. | 20:17 |
| mordred | corvus: yeah - I think probably just smallerifying that would be nice - the header there is way bigger than any other text | 20:18 |
| mordred | corvus: yeah - becuase if you wanted to see if more info - you can always click details | 20:18 |
| mordred | but you probably never want to see an empty list of numbers | 20:18 |
| corvus | the only thing i worry about in making it the same size as the other headers is whether you can tell when you switch from item 0 to item 1 | 20:19 |
| mordred | corvus: I think we can probablky leave it for a bit and mull it - it's not offensively big when there are actual results | 20:19 |
| corvus | we could break out the old <hr> horizontal rule :) | 20:19 |
| mordred | corvus: what about <blink> | 20:20 |
| mordred | corvus: as long as it's not an embedded sound file that plays when the page opens | 20:20 |
| corvus | we totally forgot to put an under construction icon on here, how will people know we're not done? | 20:21 |
| fungi | you're done when the text catches fire and the skulls spin properly | 20:23 |
| corvus | this form of output is making me think of a bunch of changes i want to make to the devstack playbook | 20:25 |
| fungi | it would be a really great job to exercise this with, since it's not shy about having a ton of tasks | 20:28 |
| corvus | oh yeah, it's one of the 3 builds i have open for testing | 20:29 |
| corvus | because i have it open, i can see it's doing some things it doesn't need to be doing :) | 20:30 |
| mordred | corvus: :) | 20:30 |
| mordred | corvus: also - have I mentioned being able to look at dashboard changes against live data is amazing? | 20:30 |
| corvus | i've never seen a complete summary of all the tasks in a playbook, all at once with one per line before | 20:30 |
| corvus | mordred: it's killer | 20:31 |
| corvus | also, ftr, i still really want to move the additionalinfo stuff farther to the right, but that's some serious css magic that i don't understand | 20:33 |
| *** jeliu_ has quit IRC | 20:59 | |
| *** tjgresha has quit IRC | 21:28 | |
| *** jamesmcarthur has quit IRC | 21:29 | |
| *** jamesmcarthur has joined #zuul | 21:29 | |
| openstackgerrit | James E. Blair proposed zuul/zuul master: Adjust results headings in build console page https://review.opendev.org/675203 | 21:30 |
| corvus | mordred: ^ | 21:30 |
| *** tjgresha has joined #zuul | 21:31 | |
| *** tjgresha has quit IRC | 21:32 | |
| mordred | corvus: yes | 21:32 |
| corvus | i just had a crazy idea | 21:32 |
| corvus | what if we went back to having one line per task, then put the result of each node in the additionalInfo section | 21:33 |
| corvus | that would probably work for maybe up to 8 nodes for a single task, which is probably okay for zuul | 21:33 |
| *** jamesmcarthur has quit IRC | 21:34 | |
| corvus | maybe 6 if the node names are longish | 21:34 |
| corvus | i guess that's not that different than before except for 2 things -- the different status values are discernable because they would be adjacent to node names, and we wouldn't have expandable child nodes underneath for the output (we'd just combine them all). but now that i imagine that, i think that tasks like "run devstack" on multiple nodes would be unusable... so i'm going to mull on that some more. | 21:36 |
| mordred | corvus: yeah - run devstack is the one that makes everything tricky | 21:39 |
| *** jamesmcarthur has joined #zuul | 21:40 | |
| *** rlandy|rover is now known as rlandy|rover|bbl | 22:22 | |
| openstackgerrit | James E. Blair proposed zuul/zuul master: Scroll log line anchor into view https://review.opendev.org/675220 | 22:23 |
| corvus | that fixes the link-to-log-line issue clarkb noticed yesterday | 22:23 |
| *** jamesmcarthur has quit IRC | 22:41 | |
| *** jamesmcarthur has joined #zuul | 22:42 | |
| *** jamesmcarthur has quit IRC | 22:47 | |
| *** armstrongs has joined #zuul | 22:50 | |
| *** jamesmcarthur has joined #zuul | 22:50 | |
| *** jamesmcarthur has quit IRC | 22:57 | |
| *** armstrongs has quit IRC | 23:00 | |
| *** michael-beaver has quit IRC | 23:14 | |
| *** tosky has quit IRC | 23:20 | |
| mnaser | so i'm hacking on builting a (somewhat) highly available zuul control plane, obviously zuul-scheduler is the tricky bit. if i use pacemaker and run 1 instance of it on any given node, is there anythign that is sticking out as a big "nono"? | 23:21 |
| mnaser | the idea is that any control plane node can disappear without the need of manual intervention | 23:22 |
| fungi | if the scheduler disappears (currently) you'll lose all builds which were queued because that's puerly in memory (for now) | 23:27 |
| fungi | er, purely | 23:27 |
| *** jamesmcarthur has joined #zuul | 23:28 | |
| fungi | also queued events/results | 23:29 |
| fungi | i'm not sure what else you would lose with a hot/cold failover of the scheduler | 23:29 |
| fungi | that might be all? | 23:29 |
| fungi | assuming all configuration was kept in sync | 23:30 |
| fungi | though the next major feature planned is to move all that state into zookeeper (and drop gearman, i think?) | 23:31 |
| *** jamesmcarthur has quit IRC | 23:33 | |
| *** igordc has quit IRC | 23:35 | |
| mnaser | fungi: ok so it will pretty much result in all jobs being lost but at least the 'service' can be restored automatically without manual intervention | 23:38 |
| mnaser | should the scheduler crash anyways | 23:38 |
| fungi | that's what i expect, purely from experience restarting the scheduler daemon | 23:42 |
| *** jamesmcarthur has joined #zuul | 23:43 | |
| mnaser | i think i might have found a dib bug with buster | 23:43 |
| mnaser | https://github.com/openstack/diskimage-builder/blob/cfba9ea79d903574afe58bbe3964c050676ee8da/diskimage_builder/lib/common-functions#L68-L72 | 23:43 |
| mnaser | seems to `wait` here forever :( | 23:43 |
| mnaser | Thinking out loud though.. I guess it would be nice if we had sometgibf similar to Jenkins shutdown mode | 23:45 |
| *** jamesmcarthur has quit IRC | 23:46 | |
| mnaser | In order to stop queueing any more jobs in a maintenance | 23:46 |
| fungi | well, zuul does have a graceful shutdown i think | 23:50 |
| fungi | but if you have jobs which run for three hours (hello tripleo) you don't necessarily want to 1. wait three hours for a scheduler restart and 2. pile up 3 hours worth of new events to then try to start jobs for after | 23:51 |
| *** mattw4 has quit IRC | 23:51 | |
| fungi | so we usually just cut our losses, use a script to create a snapshot of what items are queued in important pipelines, and then reenqueue them after | 23:52 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!
