| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: sql: add default_url column https://review.openstack.org/583066 | 00:16 |
|---|---|---|
| mordred | tristanC: I think the tenantname singleton service is the way ... and yes, please work on it if you have a chance | 00:24 |
| mordred | tristanC: I was also looking at https://angular.io/guide/router#lazy-loading-route-configuration - which isn't really quite right but has some ideas | 00:24 |
| mordred | tristanC: also https://angular.io/guide/router#child-routing-component | 00:24 |
| mordred | tristanC: which talks about how to have components that have their own roting | 00:25 |
| mordred | it still seems liek they take static input though | 00:25 |
| tristanC | mordred: yeah, i think we need to start having a context object to store a global state. singleton service seems like it would do the trick | 00:26 |
| tristanC | alright, then i'll work on that now | 00:27 |
| mordred | tristanC: I *think* if we can figure out the route stuff - we can put some context into the ActivatedRoute object (I saw something about that in the child-routing section) | 00:29 |
| mordred | tristanC: but yes- also - singleton service that handles this is ++ | 00:29 |
| tristanC | i'll try to implement an ApiInfoService | 00:30 |
| mordred | tobiash: https://review.openstack.org/#/c/583062/ <-- you should be able to install a few less apk packages | 00:34 |
| *** jiapei has joined #zuul | 02:10 | |
| *** rlandy has quit IRC | 02:28 | |
| *** _ari_ has quit IRC | 02:30 | |
| *** weshay has quit IRC | 02:30 | |
| *** _ari_ has joined #zuul | 02:31 | |
| *** pabelanger has quit IRC | 02:31 | |
| *** _ari_ has quit IRC | 02:31 | |
| *** _ari_ has joined #zuul | 02:31 | |
| *** pabelanger has joined #zuul | 02:32 | |
| *** weshay has joined #zuul | 02:32 | |
| *** sshnaidm|bbl has quit IRC | 03:49 | |
| *** sshnaidm|bbl has joined #zuul | 03:50 | |
| *** sshnaidm|bbl has quit IRC | 03:56 | |
| *** sshnaidm|bbl has joined #zuul | 03:57 | |
| jhesketh | ianw: are you going to the openinfra day in sydney next month? | 05:13 |
| *** nchakrab has joined #zuul | 05:27 | |
| *** nchakrab has quit IRC | 05:38 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: dashboard: make ZuulService load info api to manage paths https://review.openstack.org/579418 | 05:44 |
| tristanC | mordred: angular is very frustrating... 579418 took way longer than expected, i don't understand how component init works and had to use await to wait for the zuul service to get the info object | 05:45 |
| *** quiquell|off is now known as quiquell | 05:45 | |
| tristanC | it's surely not ideal, but it fixed the multi-tenant view bug and it should work as expected (e.g. white label shows status page, multi-tenant shows tenant-list if no tenant name is in the path) | 05:46 |
| *** nchakrab has joined #zuul | 06:06 | |
| *** nchakrab has quit IRC | 06:31 | |
| tristanC | corvus: Shrews: it seems like we are hitting quota mis-calculation fixed by https://review.openstack.org/536930, any chance to get a nodepool release soon? | 06:35 |
| *** nchakrab has joined #zuul | 06:37 | |
| tristanC | is there a release frequency or schedule already discussed? | 06:37 |
| *** chkumar|ruck has joined #zuul | 06:44 | |
| *** nchakrab has quit IRC | 07:09 | |
| ianw | jhesketh: i was thinking of, are you? | 07:29 |
| *** hashar has joined #zuul | 07:38 | |
| *** jluk has joined #zuul | 07:39 | |
| jluk | Hi there. Is there any way to manually enqueue a zuul buildset with a custom variable/param? I know there's nothing like this in the CLI client enqueue[-ref] command, but maybe in the protocol? | 07:47 |
| *** sshnaidm|bbl is now known as sshnaidm|rover | 07:56 | |
| openstackgerrit | Fabien Boucher proposed openstack-infra/zuul master: Add tenant yaml validation option to zuul client https://review.openstack.org/574265 | 08:00 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: dashboard: make ZuulService load info api to manage paths https://review.openstack.org/579418 | 08:02 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: web: add selenium tests for the status page https://review.openstack.org/579039 | 08:02 |
| *** electrofelix has joined #zuul | 08:10 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: sql: add default_url column https://review.openstack.org/583066 | 08:20 |
| tobiash | jluk: no, all variables parameters are defined in the job descriptions | 09:06 |
| *** nchakrab has joined #zuul | 10:00 | |
| *** nchakrab_ has joined #zuul | 10:01 | |
| *** nchakrab has quit IRC | 10:04 | |
| jluk | tobiash: ok, expected that. So, let the hacking begin :) | 10:21 |
| odyssey4me | Hi folks - is it possible for nodepool to make use of different private keys for authentication between providers? ie use private_key1 for provider 1, and private_key2 for provider 2 | 10:28 |
| odyssey4me | There is 'key-pair' which is per label, but that only defines the name for the openstack driver when creating an instance... so it defines the target public key. | 10:30 |
| odyssey4me | I guess it'd be possible to split launcher configs so that the launcher for that region has a different default id_rsa.pub, but that's not so ideal. | 10:31 |
| tobiash | odyssey4me: nodepool only does a ssh-keyscan | 10:31 |
| tobiash | odyssey4me: it doesn't do logins | 10:32 |
| odyssey4me | ah, ok - so it doesn't matter | 10:32 |
| odyssey4me | heh, makes sense | 10:32 |
| tobiash | odyssey4me: the zuul-executor knows the private key (and I think it's just one you can use atm) | 10:33 |
| odyssey4me | tobiash: yeah, makes more sense thanks - we're only using nodepool, not zuul (we use jenkins instead) | 10:38 |
| *** sambetts|afk is now known as sambetts | 10:59 | |
| *** rcarrill1 is now known as rcarrillocruz | 11:02 | |
| *** lennyb has quit IRC | 11:23 | |
| *** nchakrab_ has quit IRC | 12:07 | |
| *** rlandy has joined #zuul | 12:28 | |
| Shrews | tristanC: there is no set release frequency, just whenever we need to release, we'll release | 12:31 |
| *** elyezer has quit IRC | 12:46 | |
| *** dkranz has quit IRC | 12:50 | |
| *** samccann has joined #zuul | 13:04 | |
| tobiash | Shrews: is there a current release blocker for nodepool? | 13:04 |
| Shrews | tobiash: i don't think so | 13:05 |
| tobiash | Shrews: was there a nodepool restart in openstack after the last fixes? | 13:05 |
| *** elyezer has joined #zuul | 13:07 | |
| Shrews | tobiash: we restarted before tagging the last release. we haven't restarted with any fixes after that (that i'm aware of) | 13:07 |
| mordred | Shrews: we should probably restart with the sdk patch applied | 13:08 |
| mordred | tristanC: great job on that patch - and sorry it was frustrating! | 13:09 |
| tristanC | Shrews: then perhaps the quota fix from tobiash call for a release? | 13:13 |
| Shrews | tristanC: maybe? i don't control the release | 13:14 |
| Shrews | i'm not opposed to one | 13:14 |
| tristanC | mordred: oh well, javascript... if you don't mind the await usage then it's good, i couldn't make observer or eventemitter work similarly | 13:15 |
| Shrews | but i think openstack infra will want to restart nodepool again with the latest changes before a release | 13:15 |
| mordred | tristanC: it doesn't bother me - I think in this case we have to await because the routing information is the one thing we can't just fill in later | 13:16 |
| tristanC | mordred: otherwise i think we need to make each component subscribe to the zuul service and proceed to futher api call in a callback | 13:16 |
| jhesketh | ianw: sorry I missed your reply. Not sure yet, but was thinking about maybe submitting a zuulv3 talk if you were interested in co-presenting? | 13:17 |
| mordred | jhesketh, ianw: speaking of talks ... what are people submitting for LCA if anything? | 13:17 |
| tristanC | mordred: which could be implemented like that in a follow-up too | 13:18 |
| mordred | tristanC: nod. well, I think this at least un-breaks us | 13:18 |
| mordred | tristanC: and lets us step back and consider the overall structure | 13:18 |
| Shrews | mordred: i don't think anyone else is interested in reviewing your shade-to-sdk patch, so maybe we should just land it by EOD today? | 13:19 |
| mordred | Shrews: yah' | 13:21 |
| mordred | Shrews: I mean- I don't blame them :) | 13:21 |
| mordred | Shrews: oh - the recheck experimental did not work | 13:22 |
| Shrews | mordred: b/c of dib issues. ianw put a fix up for dib pylint | 13:22 |
| Shrews | mordred: we could either wait for that to land, or make a dependency | 13:22 |
| mordred | ah - ok | 13:23 |
| Shrews | mordred: https://review.openstack.org/583068 | 13:23 |
| mordred | Shrews: nod. I think we've banged on it pretty decently - if we restart it and things fall apart, we can revert | 13:25 |
| jhesketh | mordred: great question.. I didn't have anything in mind yet. You should totally submit and come though ;-) | 13:27 |
| openstackgerrit | Merged openstack-infra/nodepool master: Replace shade and os-client-config with openstacksdk. https://review.openstack.org/572829 | 13:38 |
| mordred | Shrews: ^^ there we go | 13:39 |
| *** dkranz has joined #zuul | 13:54 | |
| *** sshnaidm|rover is now known as sshnaidm|afk | 14:06 | |
| *** nchakrab has joined #zuul | 14:11 | |
| *** nchakrab_ has joined #zuul | 14:14 | |
| *** nchakrab has quit IRC | 14:17 | |
| Shrews | mordred: we'll need changes to the nodepool logging configs for sdk (so we don't forget) | 14:22 |
| fungi | what sort of deprecation/announcement period do we need for https://review.openstack.org/579434 (if any?) | 14:23 |
| mordred | fungi: I'm not sure - but I'm also not sure this will work as intended | 14:25 |
| mordred | because you can't pass a secret defined in one repo to a job defined in another repo | 14:26 |
| mordred | so people wanting to do this would need to make their own job that uses that role | 14:27 |
| fungi | mordred: ahh, right can't make a variant of the job from zuul-jobs with the secret as an additional parameter | 14:27 |
| mordred | I think it might be worth pondering this use case and this specific job though and seeing what we think about it | 14:27 |
| mordred | fungi: yah - but I think this shows a clear usecase where such a behavior would be desirable from a user | 14:27 |
| fungi | well, the alternative you had suggested was creating an account on rtd for the zuul server itself and then letting projects use the job with the embedded credential (and they'd have to grant the corresponding account access to their rtd project, similar to how our upload to pypi job works) | 14:28 |
| *** gtema has joined #zuul | 14:34 | |
| *** quiquell is now known as quiquell|off | 14:35 | |
| tristanC | mordred: same for upload-pypi jobs, i asked on http://lists.zuul-ci.org/pipermail/zuul-discuss/2018-January/000212.html | 14:36 |
| mordred | fungi: yes - I think we could do that - but it seems like a legit case of desired reuse - and of course corvus has already laid out the implications we'd need to satisfy in order to make such a thing | 14:40 |
| *** acozine1 has joined #zuul | 14:44 | |
| *** nchakrab has joined #zuul | 14:49 | |
| *** gtema has quit IRC | 14:50 | |
| *** nchakra__ has joined #zuul | 14:50 | |
| *** nchakrab_ has quit IRC | 14:53 | |
| *** nchakrab has quit IRC | 14:53 | |
| *** sshnaidm|afk is now known as sshnaidm|rover | 14:54 | |
| *** nchakrab has joined #zuul | 15:06 | |
| *** nchakrab_ has joined #zuul | 15:07 | |
| *** nchakrab_ has quit IRC | 15:08 | |
| *** nchakrab_ has joined #zuul | 15:08 | |
| *** nchakra__ has quit IRC | 15:09 | |
| *** nchakrab has quit IRC | 15:10 | |
| *** myoung is now known as myoung|biaf | 15:27 | |
| *** GonZo2000 has joined #zuul | 15:35 | |
| *** GonZo2000 has joined #zuul | 15:35 | |
| *** jiapei has quit IRC | 15:49 | |
| *** harlowja has joined #zuul | 16:18 | |
| corvus | mordred: something like "allow a secret to be used by parent-job playbooks, but not children" ? | 16:18 |
| *** myoung|biaf is now known as myoung | 16:19 | |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Docs: add a cross-ref in secrets https://review.openstack.org/583301 | 16:21 |
| mordred | corvus: yeah. or maybe *hand wavey* only if the secret is attached to the job in a variant? | 16:21 |
| corvus | mordred: or project-pipeline variant? | 16:22 |
| mordred | yeah | 16:22 |
| mordred | I don't know how possible it is to distinguish - but "passing a secret as an 'argument' to a job" feels more like "I added a secret in a variant" than "I added a secret in a child job" | 16:23 |
| mordred | corvus: I'm wildly gesticulating though | 16:23 |
| corvus | mordred: i agree this is worth thingking about and there may be opportunities for improvement. i have many things rattling around in my head right now, so i may let the subconcious work on this one for a few days if further discussion isn't urgent. | 16:23 |
| mordred | corvus: I do not think it is urgent | 16:23 |
| corvus | mordred: yeah, i grok the feeling you are trying to convey. | 16:23 |
| corvus | mordred: can you look at https://review.openstack.org/581888 and https://review.openstack.org/581909 when you have a sec? | 16:41 |
| mordred | corvus: for you, I've got 5 | 16:42 |
| corvus | mordred: are the pbrx changes ready for final review? | 16:43 |
| mordred | corvus: yes. | 16:44 |
| mordred | corvus: I hit recheck on https://review.openstack.org/#/c/582396/ because I added a print to the end of hte pbrx job that will output the images built | 16:44 |
| mordred | but that's cosmetic | 16:44 |
| *** hashar is now known as hasharAway | 16:45 | |
| corvus | mordred: thanks; i'll try to review them today | 16:45 |
| mordred | woot! | 16:45 |
| mordred | corvus: it should also be pretty easy to build locally if you want to poke at the results - 'pbrx --debug build-images' in your zuul directory should Just Work | 16:46 |
| *** nchakrab has joined #zuul | 16:48 | |
| *** nchakrab_ has quit IRC | 16:50 | |
| *** tobasco is now known as tobasco_away | 16:54 | |
| mordred | corvus: you may want to add https://review.openstack.org/#/c/582732 to your list (it had a bad topic earlier) | 16:56 |
| corvus | mordred, tobiash: i have left some general ruminating comments on https://review.openstack.org/582794 i'm interested in your thoughts on that. | 16:57 |
| corvus | mordred: ack, thx | 16:57 |
| openstackgerrit | Merged openstack-infra/zuul master: zk: retry initial zookeeper connection attempts https://review.openstack.org/576048 | 16:57 |
| openstackgerrit | Merged openstack-infra/zuul master: dashboard: make ZuulService load info api to manage paths https://review.openstack.org/579418 | 16:57 |
| openstackgerrit | Merged openstack-infra/zuul master: scheduler: fix enqueue event to use canonical project name https://review.openstack.org/580040 | 16:57 |
| *** nchakrab has quit IRC | 17:02 | |
| *** nchakrab has joined #zuul | 17:02 | |
| *** acozine1 has quit IRC | 17:02 | |
| openstackgerrit | Merged openstack-infra/zuul master: Stop publishing docs to docs.openstack.org https://review.openstack.org/582323 | 17:04 |
| openstackgerrit | Merged openstack-infra/zuul master: Fix line numbers in job repr https://review.openstack.org/581909 | 17:04 |
| openstackgerrit | Merged openstack-infra/zuul master: Log cpu times of ansible executions https://review.openstack.org/582713 | 17:14 |
| *** jluk has quit IRC | 17:19 | |
| openstackgerrit | Merged openstack-infra/zuul-sphinx master: Stop publishing to docs.openstack.org https://review.openstack.org/582326 | 17:20 |
| openstackgerrit | Merged openstack-infra/nodepool master: Stop publishing docs to docs.openstack.org https://review.openstack.org/582325 | 17:28 |
| *** goern has joined #zuul | 17:31 | |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul-jobs master: DNM Set pypi mirror to use 8080 instead of 80 https://review.openstack.org/583317 | 17:39 |
| *** sambetts is now known as sambetts|afk | 17:41 | |
| *** myoung is now known as myoung|lunch | 17:50 | |
| *** harlowja has quit IRC | 17:54 | |
| *** GonZo2000 has quit IRC | 17:59 | |
| *** ianychoi has quit IRC | 18:14 | |
| clarkb | looking at slowness of zuul to enqueue changes from github, I'm beginning to wonder if part of it is a depends on to Gerrit | 18:15 |
| clarkb | specifically https://github.com/kata-containers/proxy/pull/74 seems to be recheckable quickly if its depends on is merged but if it isn't merged then it takes forever (up to several hours) to show up in the zuul status at all | 18:15 |
| clarkb | but I've had a hard tiem digging through logs to see where it is blocking so unsure if that is just coincidence or related | 18:16 |
| *** electrofelix has quit IRC | 18:16 | |
| *** harlowja has joined #zuul | 18:32 | |
| *** sshnaidm|rover has quit IRC | 18:37 | |
| *** sshnaidm has joined #zuul | 18:39 | |
| *** samccann has quit IRC | 18:39 | |
| *** sshnaidm is now known as sshnaidm|rover | 18:44 | |
| *** myoung|lunch is now known as myoung | 18:47 | |
| openstackgerrit | David Shrewsbury proposed openstack-infra/nodepool master: Fix devstack plugin for using openstacksdk https://review.openstack.org/583338 | 18:51 |
| Shrews | mordred: that ^^ bit was missed in the sdk stuff. Not entirely sure if the whole LIBS_FROM_GIT stuff still applies in zuulv3 world? | 18:53 |
| mordred | Shrews: we can probalby just delete that function - since sdk is already a part of base devstack installs | 18:55 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/nodepool master: Fix devstack plugin for using openstacksdk https://review.openstack.org/583338 | 18:55 |
| Shrews | mordred: k | 18:55 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/nodepool master: Fix devstack plugin for using openstacksdk https://review.openstack.org/583338 | 18:56 |
| *** eventingmonkey has quit IRC | 18:57 | |
| *** eventingmonkey has joined #zuul | 18:58 | |
| openstackgerrit | David Shrewsbury proposed openstack-infra/nodepool master: Fix plugin and examples for using openstacksdk https://review.openstack.org/583338 | 19:03 |
| corvus | Shrews: libs_from_git is set automatically for required-projects by the devstack job | 19:14 |
| corvus | Shrews: so since openstackdsk is in required-projects in func-src-base, i think it should work as we expect (it will install openstacksdk from source always) | 19:15 |
| Shrews | corvus: i figured as much, but i wasn't sure how to verify if we still needed that sh func | 19:16 |
| clarkb | devstack or your devstack plugin must know to install it from source though | 19:16 |
| clarkb | (after checking libs from git, not sure if done for sdk) | 19:16 |
| corvus | true -- so yeah, devstack would have to support installing openstackdk from git | 19:16 |
| corvus | _install_lib_from_source "openstacksdk" | 19:17 |
| corvus | that looks promising :) | 19:17 |
| corvus | so i think we're good | 19:18 |
| corvus | if you really wanted to be sure, you could push up a change to break openstacksdk and then depends-on it from a change to nodepool | 19:18 |
| *** samccann has joined #zuul | 19:26 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul-jobs master: Add role to ensure per-node output dirs exist https://review.openstack.org/511823 | 19:41 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul-jobs master: Add role to fetch zuul logs from nodes https://review.openstack.org/583346 | 19:41 |
| SpamapS | Hey has anybody worked on a manual-triggering via web interface thus far? | 19:50 |
| SpamapS | I'm going through the things Jenkins does for us to try and migrate as much off Jenkins and that's one thing Zuul isn't so great at. | 19:50 |
| SpamapS | (we have jobs where we plug in things like downtime windows and it goes off and runs automation to notify users and modify configs appropriately) | 19:51 |
| mordred | SpamapS: yeah - I'd be kind of concerned about adding support for the full jenkins parameterized build concept - because it's kind of a thing that stems from a very different worldview jenkins has | 19:56 |
| mordred | SpamapS: that said, I think tristanC and/or mhu have wanted some web-based triggering and/or worked towards such a thing | 19:56 |
| mhu | mordred, SpamapS indeed we did! https://review.openstack.org/#/c/576907/ | 19:57 |
| SpamapS | The closest I have to it is adding a branch to an existing repo and merging empty commits. :-P | 19:57 |
| SpamapS | ohh yeah that's nice | 19:58 |
| SpamapS | and autohold interface too | 19:58 |
| SpamapS | yay | 19:58 |
| mordred | SpamapS: have you considered putting downtime windows into a yaml file or somethign similar so that you can propose a change to announce a downtime and then land it when people agree abotsuch a thing? | 19:58 |
| SpamapS | I have all of that yes. | 19:58 |
| mordred | kk | 19:59 |
| SpamapS | But these are ephemeral things. | 19:59 |
| mordred | sure | 19:59 |
| SpamapS | We'd have a repo full of thousands of pointless yaml files if we did it for everything. | 19:59 |
| mordred | just saying - if I have to click in a web browser and type text I always consider that a more systemic failure myself | 19:59 |
| mordred | but if it works for you, then it works for you :) | 19:59 |
| SpamapS | So we have a bot that does most of it without anyone thinking, in dev | 20:00 |
| SpamapS | but even after that is done, there will be times where we need to override what it does. | 20:00 |
| SpamapS | like not everything gets done with a perfect automation, sometimes you do need people to plug in overrides and teaching them yaml syntax vs. handing them a web form is not a great trade off. | 20:01 |
| mordred | heh | 20:01 |
| SpamapS | We actually have moved a lot to our slack bot too | 20:02 |
| SpamapS | so you just @Bot do this thing with this data | 20:02 |
| SpamapS | and that's easier than yaml syntaxing :) | 20:02 |
| SpamapS | so another option is just being able to feed job parameters in via *anything* that isn't git | 20:02 |
| SpamapS | but we could just make the bot make yaml and land it in git. | 20:02 |
| mordred | I guess I just don't understand what 'job parameters' are | 20:02 |
| SpamapS | The most common case I have, it's a particular change order ID and an override scheduled time. | 20:03 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul-jobs master: Add role to fetch zuul logs from nodes https://review.openstack.org/583346 | 20:04 |
| mordred | and then what happens when you trigger a job with a change order id and a time? | 20:04 |
| SpamapS | the automation interrogates the change order system to find the server, asks openstack what resources are on the server, then notifies the owners of those resources of the scheduled downtime and modifies the change order appropriately. | 20:04 |
| mordred | gotcha. so sort of like passing two -e values to an ansible playbook | 20:04 |
| SpamapS | also it disables the HV in openstack | 20:04 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: web: add tenant-scoped autohold, enqueue (change, ref) https://review.openstack.org/576907 | 20:04 |
| mhu | SpamapS, ^ needed a small rebase | 20:05 |
| SpamapS | mordred: yeah, it's just two inputs to a script | 20:05 |
| SpamapS | mhu: cool, I'll keep track. :) | 20:05 |
| pabelanger | ianw: updated commit message | 20:05 |
| SpamapS | mordred: and that really is the base level thing I want. Arbitrary inputs to script. I *could* totally make a post job in a zuul.yaml and just have people land changes that set those values. | 20:06 |
| corvus | and to clarify, i think some version of mhu's approach is likely to land. but it does not include 'parameters' -- that is something that we've restisted pretty strongly as it goes pretty hard against the grain of zuul, where all of the state is in git and things are repeatable. | 20:06 |
| SpamapS | But that's just not a great experience. | 20:06 |
| mordred | SpamapS: so - I'll have to defer to corvus on his thoughts - but to me that sounds dangerously like turning zuul in to a competitor in the space of "arbitrary automation engine" ... yeah, what corvus said | 20:06 |
| *** nchakrab has quit IRC | 20:07 | |
| SpamapS | Well I mean I can wrap a zuul.yaml editor around github. Just wondering if it's something others have run into. | 20:07 |
| SpamapS | The problem with *not* making zuul an arbitrary automation engine is that now I have two ways I have to run automation. | 20:07 |
| SpamapS | One in test, and another in live. And that is sometimes very frustrating because the subtle differences bite us. | 20:08 |
| SpamapS | Whether it's a different version of ansible, or git repos in slightly different states.. We're in that situation now, and we are fine, but occasionally we find ourselves in a corner crying. | 20:08 |
| mordred | SpamapS: I'm not saying the use-case isn't valid - just that being complimentary to other things in a similar space lke AWX or Spinnaker is our current position in the ecosystem - and I'd rather stay friendly with those projects than competitive with them | 20:09 |
| SpamapS | Yeah, we've got something AWX-like | 20:09 |
| mordred | SpamapS: so the thing you're talking about is one where the ramifications may be fairly broad and it would need some careful thought as to what it means for what we're doing | 20:09 |
| SpamapS | which is, a slack bot / eventing engine + ARA | 20:09 |
| corvus | apparently awx is now called 'tower' | 20:09 |
| mordred | oh? the split didn't take? | 20:10 |
| mhu | corvus, mordred I get your reasons against parameterized jobs but from users feedback, we're seeing the same requests as SpamapS | 20:10 |
| corvus | https://docs.ansible.com/ansible/latest/reference_appendices/tower.html | 20:10 |
| SpamapS | And running the same playbook that ran in zuul, through that, has some serious problems.. not the least of which being that watching Ansible run in realtime in ARA *sucks* | 20:10 |
| SpamapS | ah tower again, ok | 20:11 |
| SpamapS | :) | 20:11 |
| mordred | corvus: oh - it was originally called awx - then it turned in to tower - then when they open sourced tower they called that awx | 20:11 |
| dmsimard | yeah, AWX is the proper name of the upstream of Tower | 20:11 |
| SpamapS | well either way.. I'd much rather just run every *ansible* piece that runs in gate, the same way in post/prod/etc., if possible. | 20:11 |
| SpamapS | Like if Spinnaker or Tower are going to come into play, I'd want to use them in gate too. :-P | 20:12 |
| SpamapS | But anyway, for parameterization, I think I understand where Zuul's at. | 20:12 |
| mordred | yah - same way in post/prod makes total sense - it's just that in check/gate there isn't really a place to provide parameters - so if the prod version si parameterized it's still different than check/gate | 20:12 |
| SpamapS | I don't mind at all having a repo that is like 'change-order-overrides' and has a post job that runs every time we land a change to the vars. | 20:12 |
| SpamapS | Would be nice to have a git log of all the change order overrides. ;) | 20:13 |
| mordred | ++ | 20:13 |
| mordred | git log turns out to be quite nice! | 20:13 |
| * corvus <3 git logs | 20:13 | |
| SpamapS | Also I do have post jobs that deploy to prod, and it works great, but they spin up a VM to run ansible, because I don't know how to do static nodesets, so, there's a lot to think about there. | 20:14 |
| SpamapS | What does work nicely is that I have a deploy-to-stage and deploy-to-prod that depends on it | 20:14 |
| SpamapS | so if the stage deploy 'splodes, prod is protected. | 20:14 |
| mordred | SpamapS: we've got some stuff coming that should help that ... and also we're about to start doing it in infra too | 20:16 |
| mordred | SpamapS: so over the next couple of months we should be able to have better conversations about post jobs deploying to prod | 20:17 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul-jobs master: Add role to fetch zuul docs from nodes https://review.openstack.org/583350 | 20:18 |
| *** hasharAway is now known as hashar | 20:24 | |
| *** dkranz has quit IRC | 20:37 | |
| pabelanger | just sent a ML post to zuul-discuss around base jobs, if people are interested. | 20:52 |
| mordred | pabelanger: I'm always interested | 20:55 |
| pabelanger | nothing major, just something I'm considering in doing for rdoproject | 20:57 |
| corvus | i read that message as "i volunteer to write the documentation for how to build and use base jobs" | 20:58 |
| pabelanger | yah, i guess I walked into that :) | 21:00 |
| *** samccann has quit IRC | 21:01 | |
| openstackgerrit | Merged openstack-infra/zuul master: Build container images using pbrx https://review.openstack.org/580160 | 21:01 |
| openstackgerrit | Merged openstack-infra/zuul master: Specify a prefix for building the images https://review.openstack.org/582396 | 21:08 |
| openstackgerrit | Merged openstack-infra/zuul master: Install less than alpine-sdk https://review.openstack.org/583062 | 21:08 |
| corvus | pabelanger: replied. also, fwiw, there's nothing magic about the name 'base'. you can set the default parent job for a tenant (it just defaults to 'base'. that's the default default.) | 21:09 |
| corvus | pabelanger: might be worth suggesting that approach in the openstack-infra context too? | 21:09 |
| pabelanger | corvus: yah, I wanted to first float the idea here. So far the largest issue we've had around base jobs, is less experienced people in rdoproject merging changes to base and hoping they work. I'll bring up the topic with openstack-infra shortly | 21:24 |
| clarkb | pabelanger: base jobs must be in trusted repos though right? | 21:25 |
| corvus | pabelanger: regardless of this, it's pretty important that anyone who has permissions to merge changes to a config repo agrees on the parameters for merging those changes. it's called "trusted" for a reason. | 21:25 |
| corvus | like, by all means, let's make things better and more convenient. but there's also an underlying issue if you can't trust the trusted repo reviewers, and that issue won't go away even if we improve the ux. it's better to fix this from a POV of "the trusted repo reviewers are inconvenienced by this" rather than "the trusted repo reviewers don't understand what they're being asked to do". | 21:27 |
| corvus | clarkb: *base* jobs must be in trusted repos. iirc, default-parent jobs can be anywhere. | 21:27 |
| corvus | (a base job is a job with no parent. the default parent for jobs which don't specify a parent is (by default) named "base". that job can be defined anywhere though) | 21:29 |
| corvus | so pabelanger's suggestion of "base-minimal" (actual base job) as the parent of "base" (counterintuitively not a base job) and using "base" as the default-parent should work. | 21:30 |
| clarkb | gotcha | 21:30 |
| corvus | as would having "base" be an actual base job and "pabelanger-base" as a child of base, and setting "default-parent: pabelanger-base" in the tenant | 21:30 |
| *** hashar has quit IRC | 21:36 | |
| openstackgerrit | Merged openstack-infra/zuul master: Fix broken templates for projects in dependent pipelines https://review.openstack.org/581888 | 22:01 |
| * mordred wants to adopt a system where one of our base jobs is called "pabelanger-base" | 22:22 | |
| *** sambetts|afk has quit IRC | 22:26 | |
| ianw | jhesketh: i did offer a infra/zuul-ish talk for the sydney infra day thing ... i framed it more as an overview of openstack's change process. i don't think they're that interested in openstack talks | 22:27 |
| *** sambetts_ has joined #zuul | 22:28 | |
| *** myoung is now known as myoung|off | 22:31 | |
| pabelanger | corvus: re: 'trusted for reason' Yup, just some growing pain on new zuulv3 users not understanding the power | 22:56 |
| *** rlandy has quit IRC | 23:24 | |
| fungi | ianw: jhesketh: i learned a few years ago to stop mentioning "openstack" in my abstacts | 23:46 |
| fungi | or i should have. i think a passing mention in my ato proposal is likely what got my open community vulnerability management talk nixed from this year's lineup | 23:46 |
| fungi | at this point conferences are afraid an "openstack" talk makes them look old and dusty | 23:47 |
| fungi | oh, openstack... that thing kubernetes replaced? | 23:47 |
| ianw | fungi: yes, this one even has the subtitle "Previously known as OpenStack Australia Day," | 23:54 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!