| sgw | evening folks, I am trying the Zuul tutorial and running into an issue with the git review / push to gerrit. I get the follow message which seems to indicate it's pushing to the wrong branch (not sure why): | 00:02 |
|---|---|---|
| sgw | ! [remote rejected] HEAD -> refs/publish/master (prohibited by Gerrit: not permitted: create) | 00:02 |
| *** mattw4 has joined #zuul | 00:02 | |
| sgw | if I push by hand to refs/for/master it seems to work, I use :git-review regularly with opendev.org with no issues, just seems to be a problem with the local tutorial gerrit. | 00:03 |
| *** gfhellma_ has joined #zuul | 00:07 | |
| *** gfhellma_ has quit IRC | 00:09 | |
| fungi | sgw: you need a newer git-review release | 00:10 |
| fungi | what version are you running? | 00:10 |
| fungi | the version of gerrit in the tutorial is much newer than what review.opendev.org runs | 00:11 |
| fungi | last gerrit release dropped support for the publish refs path, and recent git-review releases are updated to adapt | 00:11 |
| * fungi figures out the minimum gr version you need, just a sec | 00:12 | |
| fungi | https://review.openstack.org/584607 is the change which fixed it, included in the 1.27.0 release | 00:14 |
| sgw | fungi: 1.26.0, will a newer version be problematic for review.opendev.org? | 00:14 |
| fungi | newer version works just fine | 00:14 |
| fungi | i use it daily | 00:14 |
| fungi | you can install latest from pypi (1.28.0) and it'll be fine with opendev's gerrit | 00:14 |
| sgw | fungi: BTW, I also had to make sure I was using pip3 for the initial installation of the tutorial, otherwise I got an error, you might call that out in your tutorial docs | 00:15 |
| fungi | good idea. which step gave you the error? | 00:15 |
| sgw | The docker-compuse up needed a new version that what was installed by default packages (Ubuntu) | 00:16 |
| fungi | this is the "quickstart" tutorial? | 00:16 |
| sgw | yes, is there a different tutorial I should work through? | 00:16 |
| fungi | nope, just making sure you weren't talking about the "zuul from scratch" tutorial | 00:17 |
| fungi | so it was the docker-compose up in the third quote block of https://zuul-ci.org/docs/zuul/admin/quick-start.html#start-zuul-containers | 00:18 |
| sgw | Ultimately I want to learn about using a docker container to build packages in a check job | 00:19 |
| sgw | yes that quote block | 00:19 |
| fungi | source code is here: https://opendev.org/zuul/zuul/blame/branch/master/doc/source/admin/quick-start.rst#L66 | 00:21 |
| fungi | what did you end up needing to do to specify python 3? | 00:22 |
| fungi | er, pip3 | 00:22 |
| sgw | yes, I ran pip3 install docker-compose | 00:23 |
| fungi | interesting. https://opendev.org/zuul/zuul/blame/branch/master/doc/source/admin/quick-start.rst#L44 had you install docker-compose from ubuntu's packages | 00:24 |
| fungi | but that was too old i guess? | 00:24 |
| fungi | now i'm starting to understand i think | 00:25 |
| fungi | what sort of error did the older (ubuntu packaged) docker-compose give you? do you happen to have it handy? | 00:25 |
| sgw | Right that exactly what happened first a ubuntu package install and then I did the pip3 install. | 00:26 |
| sgw | Right at the top of my scroll back with all the other logging going on! | 00:27 |
| sgw | TypeError: build() got an unexpected keyword argument 'stream' | 00:27 |
| sgw | Google was able to help me figure it out. | 00:27 |
| fungi | we try to test the quick-start steps in our ci, and install docker-compose from distro packages it seems https://opendev.org/zuul/zuul/src/branch/master/playbooks/quick-start/run.yaml#L10-L15 | 00:28 |
| fungi | the job definition is at https://opendev.org/zuul/zuul/src/branch/master/.zuul.yaml#L85-L96 and doesn't specify a nodeset, so is presumably running on ubuntu-bionic (18.04 lts, opendev's default nodeset) | 00:30 |
| fungi | what ubuntu release are you running on? | 00:30 |
| *** mattw4 has quit IRC | 00:32 | |
| sgw | 18.04.2 LTS | 00:32 |
| fungi | okay, that's what we're testing it on | 00:32 |
| fungi | i wonder why we don't hit it there | 00:32 |
| sgw | no idea, but I did hit it and the package was older. | 00:34 |
| sgw | I might hit you up with more questions tomorrow if that's OK, I am off to tear up an old deck | 00:34 |
| fungi | docker-compose 1.17.1 is what's currently in bionic anyway | 00:34 |
| fungi | okay, great! have fun with the deck | 00:35 |
| mnaser | so trying to rotate credentials here https://review.opendev.org/#/c/665917/ -- it seems like zuul is unhappy because it doesn't match the secret in the other branches | 00:48 |
| mnaser | now ideally we would want to change it everywhere.. | 00:48 |
| mnaser | and it isn't really a branch-specific secret either, so i'm not sure if maybe this should live in openstack/project-config ... i dunno | 00:49 |
| fungi | i wonder if zuul should simply ignore secrets defined on a stable branch if they're also defined in master? | 00:51 |
| fungi | there's presumably some reason we added that safety catch though | 00:51 |
| fungi | deleting the copies from the stable branches should work, but obviously that makes for more work after future branching events | 00:52 |
| mnaser | yeah, in the future i assume this will just happen again on the next branch i guess | 00:52 |
| mnaser | i could go and drop them from all stable branches but just trying to future-proof this from a zuul perspective | 00:53 |
| fungi | code comments here lend some explanation: https://opendev.org/zuul/zuul/src/branch/master/zuul/model.py#L3610-L3628 | 00:56 |
| fungi | looks like it's similarly picky about divergent nodeset definitions on different branches | 00:56 |
| mnaser | yeah that explains the exact behaviour here, hmm | 00:57 |
| mnaser | i guess this was a conscious decision | 00:57 |
| fungi | it may be we were just being overly cautious, and didn't consider the implications on replacing those values in branched projects | 00:57 |
| fungi | https://review.openstack.org/535501 is the change which added that logic | 00:59 |
| fungi | there's a fairly extensive commit message with it | 00:59 |
| clarkb | 01:00 | |
| clarkb | oops | 01:00 |
| mnaser | fungi: the commit message contains a clear "We do not support different values for the same secret name on different branches." | 01:00 |
| mnaser | so i guess we should clean that up | 01:00 |
| fungi | yeah, so prior to that change we didn't support defining the same secret on multiple branches at all | 01:01 |
| fungi | that change made it possible to branch projects with secrets in them, because it became obvious that was a workflow people had | 01:01 |
| fungi | it may be a defensible position that zuul should allow (and ignore) divergent secret definitions from what's on master and still ignore the non-master copies | 01:03 |
| mnaser | yeah, but i think this seems reasonable now to avoid huge amonuts of confusion | 01:03 |
| mnaser | https://review.opendev.org/#/c/665917/ is now ok with a depends-on the stable branch that has the secret defined | 01:04 |
| fungi | though worth thinking about what happens when you have secret-using jobs on branches master, foo and bar and where all three differ and then you remove the secret from master | 01:04 |
| mnaser | which means things are working as planned(tm) | 01:04 |
| *** igordc has quit IRC | 01:16 | |
| *** mattw4 has joined #zuul | 03:00 | |
| *** bhavikdbavishi has joined #zuul | 03:13 | |
| *** bhavikdbavishi1 has joined #zuul | 03:18 | |
| *** bhavikdbavishi has quit IRC | 03:20 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 03:20 | |
| *** bhavikdbavishi has quit IRC | 04:24 | |
| *** raukadah is now known as chandankumar | 04:31 | |
| *** bhavikdbavishi has joined #zuul | 04:37 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Evaluate CODEOWNERS settings during canMerge check https://review.opendev.org/644557 | 04:54 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Optionally support mitogen for job execution https://review.opendev.org/657024 | 05:06 |
| *** irclogbot_1 has quit IRC | 05:30 | |
| *** mgoddard has quit IRC | 05:31 | |
| *** irclogbot_3 has joined #zuul | 05:32 | |
| *** mgoddard has joined #zuul | 05:33 | |
| *** spsurya has joined #zuul | 06:16 | |
| *** threestrands has joined #zuul | 06:27 | |
| *** gtema has joined #zuul | 06:29 | |
| *** themroc has joined #zuul | 07:04 | |
| *** igordc has joined #zuul | 07:05 | |
| *** igordc has quit IRC | 07:10 | |
| *** hashar has joined #zuul | 07:26 | |
| *** saneax has joined #zuul | 07:38 | |
| *** jpena|off is now known as jpena | 07:43 | |
| *** threestrands has quit IRC | 07:59 | |
| *** themroc has quit IRC | 08:10 | |
| *** themroc has joined #zuul | 08:11 | |
| *** pcaruana has quit IRC | 08:27 | |
| *** pcaruana has joined #zuul | 08:45 | |
| *** arxcruz is now known as arxcruz|brb | 08:47 | |
| *** mattw4 has quit IRC | 08:59 | |
| *** hashar has quit IRC | 09:23 | |
| *** hashar has joined #zuul | 09:24 | |
| *** hashar has quit IRC | 09:27 | |
| fbo | corvus: Pagure driver merged \o/ thanks :) ! | 09:30 |
| *** igordc has joined #zuul | 09:37 | |
| *** yolanda has quit IRC | 09:41 | |
| badboy | guys, is it possible to show IP or hostname in the Nodes tabs of static-vms? | 09:43 |
| *** electrofelix has joined #zuul | 10:03 | |
| *** electrofelix has quit IRC | 10:03 | |
| *** electrofelix has joined #zuul | 10:09 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Dequeue a change by its buildset UUID https://review.opendev.org/666258 | 10:12 |
| *** gtema has quit IRC | 10:14 | |
| *** gtema has joined #zuul | 10:15 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Dequeue a change by its buildset UUID https://review.opendev.org/666258 | 10:21 |
| *** sshnaidm is now known as sshnaidm|afk | 10:24 | |
| *** saneax has quit IRC | 10:26 | |
| *** saneax has joined #zuul | 10:26 | |
| *** sanjayu_ has joined #zuul | 10:34 | |
| *** sanjayu_ has quit IRC | 10:34 | |
| *** sanjayu_ has joined #zuul | 10:35 | |
| *** saneax has quit IRC | 10:35 | |
| *** sanjayu__ has joined #zuul | 10:37 | |
| *** sanjayu_ has quit IRC | 10:40 | |
| Shrews | badboy: No reason that info can't be added (it's stored in zookeeper, where that data comes from). That info probably isn't as useful for the openstack driver (which is probably why it isn't there). But for the static driver, that info makes sense to have. | 10:53 |
| badboy | Shrews: I only have static-vms at the moment and that's why I'm asking about it :) | 10:54 |
| * Shrews reads email about dynamic nodesets based on user location... scratches his head | 11:03 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Dequeue a change by its buildset UUID https://review.opendev.org/666258 | 11:06 |
| *** yolanda has joined #zuul | 11:17 | |
| *** jpena is now known as jpena|lunch | 11:19 | |
| openstackgerrit | Jean-Philippe Evrard proposed zuul/zuul master: Expose ansible_date_time instead of date_time https://review.opendev.org/666268 | 11:25 |
| openstackgerrit | Jean-Philippe Evrard proposed zuul/zuul master: Expose ansible_date_time instead of date_time https://review.opendev.org/666268 | 11:30 |
| *** arxcruz|brb is now known as arxcruz | 11:53 | |
| *** rfolco has joined #zuul | 11:54 | |
| *** gtema has quit IRC | 12:00 | |
| tobiash | we never had user based things in jobs and I'm not really sure if that's desirable | 12:06 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: web: add tenant and project scoped, JWT-protected actions https://review.opendev.org/576907 | 12:16 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Dequeue a change by its buildset UUID https://review.opendev.org/666258 | 12:18 |
| *** gtema has joined #zuul | 12:21 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Allow operator to generate auth tokens through the CLI https://review.opendev.org/636197 | 12:21 |
| *** rlandy has joined #zuul | 12:27 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Zuul CLI: allow access via REST https://review.opendev.org/636315 | 12:28 |
| *** sshnaidm|afk is now known as sshnaidm | 12:29 | |
| mnaser | i have a bit of a generic question. is there a workflow that's recommended when using zuul for a periodic task that's not necessarily tied to a specific repo? | 12:29 |
| *** jpena|lunch is now known as jpena | 12:29 | |
| mnaser | for example: we need to run a periodic task that does mirroring of docker images.. it isn't tied to a code.. or commit or anything | 12:30 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Add Authorization Rules configuration https://review.opendev.org/639855 | 12:32 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Web: plug the authorization engine https://review.opendev.org/640884 | 12:35 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Zuul Web: add /api/user/authorizations endpoint https://review.opendev.org/641099 | 12:36 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: authentication config: add optional token_expiry https://review.opendev.org/642408 | 12:36 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: [WIP] admin REST API: zuul-web integration https://review.opendev.org/643536 | 12:41 |
| fungi | mnaser: perhaps a glib answer, but the configuration which defines that task needs to live in a git repository, so couldn't you say the task is tied to it? | 12:46 |
| mnaser | fungi: i guess that's what i ended up concluding too as well, i think that's the path i'll do, the odd thing is how to self-test that change when its meant to be a periodic job (but i guess thats why zuul enqueue is a thing) | 12:46 |
| fungi | with enough abstraction you could run that in the check/gate pipelines in an incomplete way, maybe "mirroring" the images to /dev/null | 12:48 |
| mnaser | yeah.. or actually have a mirror target of a locally deployed registry too | 12:48 |
| fungi | or to a local directory which then gets scooped up during log collection so that you can confirm it got the right things | 12:48 |
| mnaser | instead of dockerhub in my case | 12:48 |
| fungi | sure, fake image repository. we already have roles to make those too | 12:49 |
| mnaser | yep, that sounds like my path to success | 12:49 |
| fungi | well, not even fake. actual image repository just not one which sticks around past the end of the buildset | 12:49 |
| fungi | though https://review.opendev.org/663119 will make that role slightly safer | 12:50 |
| mnaser | oh its nice there's an actual password | 12:51 |
| mnaser | because id liek to simulate the login too | 12:51 |
| mnaser | this is aesome | 12:51 |
| fungi | yep | 12:51 |
| mnaser | is there any sample jobs i can eye (a project should be enough to point me to) | 12:52 |
| fungi | there are indeed. zuul's image build jobs are using this already | 12:53 |
| mnaser | cool, ill look at zuul/zuul then :> | 12:53 |
| fungi | (for building our zuul docker images) | 12:53 |
| fungi | lmk if you have any questions and i'll try to answer them, though some of this is still slightly voodoo to me | 12:54 |
| fungi | just because i haven't spent enough time digging into the docker protocols | 12:54 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Dequeue a change by its buildset UUID https://review.opendev.org/666258 | 12:56 |
| *** yolanda has quit IRC | 13:13 | |
| *** yolanda has joined #zuul | 13:17 | |
| mnaser | ok i've ben struggling with this for way too long.. | 13:20 |
| mnaser | ERROR! the role 'add-build-sshkey' was not found | 13:20 |
| mnaser | i do have roles: - zuul: zuul/zuul-jobs in there though | 13:20 |
| mnaser | (and i have zuul/zuul-jobs defined as an untrusted job) | 13:22 |
| pabelanger | mnaser: have you added the role to your (base) job? | 13:24 |
| mnaser | pabelanger: not to the base (tbh i am still using zuul-base-jobs right now) but i added it to the job that im running | 13:24 |
| pabelanger | mnaser: eg: https://github.com/ansible/project-config/blob/master/zuul.d/jobs.yaml#L33 | 13:25 |
| pabelanger | I am guessing you are missing that | 13:25 |
| mnaser | oh i have that but without opendev.org/ | 13:25 |
| pabelanger | mnaser: is the job you are using it trusted or untrusted? | 13:25 |
| mnaser | i think that might be it | 13:25 |
| mnaser | its an untrusted job | 13:25 |
| pabelanger | mnaser: no, opendev.org is option, that is just the connection name I have | 13:25 |
| mnaser | ah okay | 13:25 |
| pabelanger | (I use full path, to help users know where it is) | 13:25 |
| mnaser | http://paste.openstack.org/show/753187/ is the job | 13:26 |
| pabelanger | mnaser: an you have a connection to opendev.org? | 13:26 |
| mnaser | and i can see zuul/zuul-jobs listed under projects in the ui (with connection=opendev and type=untrusted) | 13:26 |
| pabelanger | okay, that should be okay | 13:26 |
| pabelanger | where is add-build-sshkey done? | 13:27 |
| pabelanger | which playbook (zuul job) | 13:27 |
| mnaser | im assuming the zuul-base-job is doing it | 13:27 |
| mnaser | but im guessing maybe thats the issue | 13:27 |
| pabelanger | you need to make sure the roles is define before or at that playbook | 13:27 |
| mnaser | i guess you cant just 'use' zuul/zuul-base-jobs even in some form of testing? | 13:27 |
| mnaser | let me get some base jobs going | 13:28 |
| mnaser | and attach those roles to the base jobs | 13:28 |
| pabelanger | mnaser: if add-build-sshkey is in base, yah the role also needs to be defined next to it | 13:31 |
| openstackgerrit | James E. Blair proposed zuul/nodepool master: WIP: new devstack-based functional job https://review.opendev.org/665023 | 13:33 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Add repl server for debug purposes https://review.opendev.org/579962 | 13:36 |
| tobiash | corvus: I have to debug a weird memleak we have on web and executor so I thought it would be a good idea to finalize this ^ | 13:37 |
| corvus | tobiash: awesome, thx! | 13:37 |
| tobiash | we currently have the problem that our executors show the same memleak we have also for zuul-web which renders filesystem caches useless after one day | 13:39 |
| corvus | tobiash: cool, i like that as a first step. i think leaving it undocumented is fine for now. also, i was thinking we might want to do "zuul-executor enable-repl" | 13:39 |
| tobiash | leading to job preparation times of up to an hour :/ | 13:39 |
| tobiash | corvus: that is a great idea :) | 13:39 |
| *** bhavikdbavishi has quit IRC | 13:45 | |
| tobiash | corvus: should we instead make it available only through the command socket? | 13:46 |
| corvus | tobiash: i do think that would be preferable | 13:47 |
| tobiash | k, I'll change that | 13:47 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: URLTrigger driver time based https://review.opendev.org/635567 | 13:47 |
| openstackgerrit | Fabien Boucher proposed zuul/zuul master: A reporter for Elasticsearch https://review.opendev.org/644927 | 13:48 |
| flaper87 | tobiash: hey, did you get the web console to work on your k8s based deployment? Mind sharing how you configured it? :) | 13:55 |
| *** sanjayu__ has quit IRC | 13:55 | |
| *** jamesmcarthur has joined #zuul | 14:05 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Get executor job params https://review.opendev.org/607078 | 14:12 |
| tobiash | flaper87: what kind of problem do you have? | 14:13 |
| flaper87 | tobiash: basically, the executor is reporting its hostname but the web can't resolve it (its hostname is basically the pod name). I may be missing one config step here. I'm not setting the executor.hostname option | 14:29 |
| flaper87 | The hostname the web service receives is correct but it can't resolve it 'cause it's not a service, it's the pod name | 14:29 |
| openstackgerrit | James E. Blair proposed zuul/nodepool master: WIP: new devstack-based functional job https://review.opendev.org/665023 | 14:30 |
| tobiash | flaper87: I'm running the executors as statefulset combined with a headless service which makes them resolvable | 14:31 |
| tobiash | then the hostnames should work | 14:31 |
| tobiash | headless service == Type clusterip AND clusterip==None | 14:32 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Add repl server for debug purposes https://review.opendev.org/579962 | 14:33 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Add command processor to zuul-web https://review.opendev.org/666307 | 14:33 |
| tobiash | corvus: now with commandsocket ^ | 14:33 |
| tobiash | (had to add it first to zuul-web) | 14:33 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Add command processor to zuul-web https://review.opendev.org/666307 | 14:34 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Add repl server for debug purposes https://review.opendev.org/579962 | 14:34 |
| flaper87 | tobiash: oh, mmh, lemme try something like that. | 14:34 |
| flaper87 | tobiash: any reason why you went with statefulset instead of deployment ? | 14:35 |
| tobiash | exactly because of this | 14:35 |
| tobiash | having individual pods resolvable is only possible with statefulsets | 14:35 |
| flaper87 | lol, good! | 14:36 |
| flaper87 | In that case, lemme change | 14:36 |
| flaper87 | it | 14:36 |
| *** gfhellma has joined #zuul | 14:45 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Add repl server for debug purposes https://review.opendev.org/579962 | 14:57 |
| *** michael-beaver has joined #zuul | 15:05 | |
| *** mattw4 has joined #zuul | 15:11 | |
| tobiash | corvus: did you see my comment on 666177? | 15:13 |
| tobiash | how urgent is this? | 15:13 |
| *** themroc has quit IRC | 15:18 | |
| openstackgerrit | Jean-Philippe Evrard proposed zuul/zuul master: Expose ansible_date_time instead of date_time https://review.opendev.org/666268 | 15:25 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Dequeue a change by its buildset UUID https://review.opendev.org/666258 | 15:39 |
| *** gfhellma has quit IRC | 15:40 | |
| mhu | hey jhesketh I think I broke your patch by attempting a rebase: https://review.opendev.org/#/c/607078/ my apologies! | 15:42 |
| corvus | tobiash: yes; that solution sounds okay, though we should sit down and see about addressing your underlying problems at some point :) | 15:43 |
| corvus | tobiash: it's not urgent | 15:43 |
| tobiash | corvus: yes, what I meant was it would be good if I could get a few days to address my underlying problems with that | 15:44 |
| tobiash | the easiest would be a config option in the diskimage that specifies the diskimage builder executable to be used | 15:44 |
| corvus | tobiash: yes, i think that's okay (but not ideal -- the software depends on DIB, not on "something that implements a DIB interface" -- when DIB is python-importable, we have every intention of switching to that, which would break your hack) | 15:46 |
| corvus | tobiash: which is why i say that at some point we should figure out how to make DIB serve your needs better | 15:46 |
| tobiash | I don't think dib can be changed to fit our windows needs | 15:46 |
| pabelanger | tobiash: I think that would be a great idea, we are use case for virt-customize for network appliances | 15:47 |
| pabelanger | (not to side track) | 15:47 |
| corvus | tobiash: sure, which probably means you need a different builder for windows | 15:47 |
| *** mattw4 has quit IRC | 15:47 | |
| corvus | tobiash: nodepool is designed to support multiple builders | 15:47 |
| *** mattw4 has joined #zuul | 15:47 | |
| tobiash | that would be the other option | 15:48 |
| tobiash | we use a custom ansible based builder that imitates the diskimage-builder cli interface | 15:48 |
| corvus | we shouldn't try to shoe-horn everything into the DIB builder | 15:48 |
| corvus | we should have an ansible builder :) | 15:48 |
| pabelanger | YES | 15:48 |
| pabelanger | Please :D | 15:48 |
| corvus | you don't have to ask me | 15:49 |
| corvus | just do it | 15:49 |
| pabelanger | (TM) | 15:49 |
| tobiash | that sounds good | 15:49 |
| corvus | this design even pre-dates zuulv3... | 15:49 |
| tobiash | we currently drive a similar way like dib (we treat a list of playbooks like the elements) | 15:50 |
| pabelanger | It is on the list of things to do, so far we are relying on https://github.com/ansible-network/network-image-builder/ for network image builds. Would be fantastic to have nodepool-builder call that directly | 15:50 |
| corvus | nodepool v2 had a multi-builder design. the fact that there wasn't a second builder merely reflects the fact that openstack hasn't needed a second builder | 15:50 |
| pabelanger | Yah | 15:50 |
| corvus | or, i should say, nodepool v0 | 15:50 |
| tobiash | I'll think about how we could integrate something like this that fits our needs | 15:50 |
| tobiash | our second use case is that we do a git-pull on our custom elements before building to make our continuous delivery easier | 15:52 |
| tobiash | otherwise we would need to rebuild/restart nodepool to update the elements | 15:52 |
| tobiash | I guess this could be solved by some pre-execution step | 15:52 |
| pabelanger | why do you need to restart? | 15:53 |
| tobiash | the elements can be either baked into the container or loaded on startup in the container | 15:54 |
| pabelanger | ah | 15:54 |
| tobiash | both involve a restart of the builder to update the elements | 15:54 |
| pabelanger | could you place them into volume? | 15:54 |
| pabelanger | or is there a downside to that? | 15:54 |
| tobiash | on the fly updating from outside is too racy | 15:55 |
| tobiash | so our solution was a wrapper that updates then executes the build | 15:55 |
| tobiash | that is simple and race free | 15:55 |
| *** mattw4 has quit IRC | 15:55 | |
| * Shrews wants to come to BMW to see tobiash's zuul setup | 15:56 | |
| Shrews | and get a free car | 15:56 |
| tobiash | and could be easily replaced by a pre-buikd step | 15:56 |
| * tobiash doesn't have a free car ;) | 15:56 | |
| tobiash | Shrews: but I'd happy to show you my zuul setup when we meet next at some conference :) | 15:57 |
| *** mattw4 has joined #zuul | 15:59 | |
| tobiash | so a plan could be 1) add diskimage-builder executable setting (optional, depending how important the venv change is), 2) create an ansible builder (might need a spec), 3) add pre-build step to facilitate preparation like in-place updates of elements before build | 16:00 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Add command processor to zuul-web https://review.opendev.org/666307 | 16:08 |
| *** spsurya has quit IRC | 16:18 | |
| *** chandankumar is now known as raukadah | 16:21 | |
| *** rfolco has quit IRC | 16:25 | |
| *** rfolco has joined #zuul | 16:26 | |
| *** mattw4 has quit IRC | 16:37 | |
| *** mattw4 has joined #zuul | 16:37 | |
| *** mattw4 has quit IRC | 16:42 | |
| *** gfhellma has joined #zuul | 16:44 | |
| *** bjackman has quit IRC | 16:45 | |
| *** jpena is now known as jpena|off | 16:45 | |
| *** gfhellma_ has joined #zuul | 16:50 | |
| *** gfhellma has quit IRC | 16:53 | |
| openstackgerrit | James E. Blair proposed zuul/nodepool master: WIP: new devstack-based functional job https://review.opendev.org/665023 | 16:56 |
| *** mgoddard has quit IRC | 17:25 | |
| *** mgoddard has joined #zuul | 17:27 | |
| *** migi has joined #zuul | 17:29 | |
| migi | Hi. I am getting some issue defining zuul job that includes role from external project. External project has a role which has include_tasks as defined here https://github.com/rhos-infra/patch-components/blob/master/roles/patch_rpm/tasks/main.yml#L96 | 17:31 |
| migi | now when I add the top level repo to the job definition via roles.zuul then it complains this particular line can not be found: Could not find or access [stripped] tasks/patch_opendaylight.yml' on the Ansible Controller. | 17:33 |
| migi | everything works fine if the project which is included does not contain the include_tasks, is this some bug in zuul or my misuse ? | 17:33 |
| pabelanger | migi: do you have an example of the failure? When you say roles.zuul, are you referring to jobs.role setting in your zuul.yaml file? | 17:36 |
| *** mattw4 has joined #zuul | 17:36 | |
| corvus | migi: relative paths in ansible can be pretty confusing; i think that include is going to be relative to the current task list, so you might need to drop "roles/patch_rpm/tasks" from it since that file is already in that directory | 17:36 |
| corvus | migi: so maybe it should just say "include_tasks: patch_component.yml" ? | 17:37 |
| migi | pabelanger: here is what I am defining http://paste.openstack.org/show/753198/ | 17:39 |
| pabelanger | I think the issue is, https://github.com/rhos-infra/patch-components/ isn't an ansible role | 17:39 |
| pabelanger | so, zuul isn't adding it to your role path | 17:40 |
| pabelanger | it is a group of roles, which I am not sure zuul supports | 17:40 |
| openstackgerrit | James E. Blair proposed zuul/nodepool master: WIP: new devstack-based functional job https://review.opendev.org/665023 | 17:40 |
| corvus | pabelanger: zuul totally supports groups of roles | 17:40 |
| migi | pabelanger: similarly the https://github.com/redhat-openstack/octario/ is added and used as multiple roles which works ok | 17:41 |
| pabelanger | okay, then ignore me :) | 17:41 |
| migi | pabelanger: the only difference between those two projects and the way it's being used is this include_tasks | 17:41 |
| corvus | pabelanger: https://zuul-ci.org/docs/zuul/user/config.html#attr-job.roles fourth paragraph "contained role" | 17:41 |
| pabelanger | corvus: thanks! | 17:42 |
| migi | corvus: ok will try that, thanks for the hint | 17:44 |
| migi | corvus: or to make it backwards compatible will try to loop over two paths, one current and one without roles/patch_rpm/tasks | 17:45 |
| corvus | migi: if omitting it works, it may be backwards compatible too | 17:46 |
| corvus | migi: (ansible often has a series of locations it searches for things, so using the "closest" path may work in both situations) | 17:47 |
| corvus | migi: you may be able to verify that with depends-on patches | 17:48 |
| migi | corvus: it's non-production version of zuul so I am fine submitting atm many changes without breaking anyone | 17:49 |
| migi | corvus: but yes depends-on is best here, thanks ! | 17:49 |
| *** hashar has joined #zuul | 17:50 | |
| *** gtema has quit IRC | 17:54 | |
| *** gtema has joined #zuul | 17:59 | |
| *** gfhellma__ has joined #zuul | 18:03 | |
| *** gfhellma_ has quit IRC | 18:06 | |
| *** electrofelix has quit IRC | 18:08 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Dequeue a change by its buildset UUID https://review.opendev.org/666258 | 18:10 |
| *** pcaruana has quit IRC | 18:11 | |
| *** jamesmcarthur has quit IRC | 18:40 | |
| *** jamesmcarthur has joined #zuul | 18:41 | |
| openstackgerrit | James E. Blair proposed zuul/nodepool master: WIP: new devstack-based functional job https://review.opendev.org/665023 | 18:45 |
| *** jamesmcarthur has quit IRC | 18:46 | |
| *** jamesmcarthur has joined #zuul | 19:04 | |
| migi | corvus: removing worked, thanks a lot ! | 19:09 |
| *** jamesmcarthur has quit IRC | 19:10 | |
| *** jamesmcarthur has joined #zuul | 19:23 | |
| *** gtema has quit IRC | 19:50 | |
| *** gfhellma__ has quit IRC | 20:07 | |
| openstackgerrit | James E. Blair proposed zuul/nodepool master: WIP: new devstack-based functional job https://review.opendev.org/665023 | 20:09 |
| *** jamesmcarthur has quit IRC | 20:13 | |
| *** jamesmcarthur has joined #zuul | 20:15 | |
| *** jamesmcarthur has quit IRC | 20:23 | |
| corvus | 3.7.0 https://opendev.org/zuul/nodepool/commit/3412764a985b511cdc6b70dc801ffdb357ec02c2 | 20:41 |
| corvus | 3.9.0 https://opendev.org/zuul/zuul/commit/217dea00251a51b23fe3d2083f965482eb71af8b | 20:41 |
| corvus | pabelanger, fungi, tobiash, mordred: ^ how's that look? | 20:42 |
| tobiash | lgtm | 20:44 |
| pabelanger | +1 | 20:44 |
| fungi | 90 commits for zuul between 3.8.1 and the proposed 3.9.0 at least some of which look like more than a patch level increase, so that seems reasonable | 20:45 |
| fungi | only 16 for nodepool between 3.6.0 and the proposed 3.7.0 but at least some stuff (at least the openshift pod provider) warrants the minor version increase | 20:47 |
| fungi | yeah, lgtm corvus! | 20:47 |
| corvus | that whole gpg agent thing that broke our ability to edit password files on bridge has also broken my ability to sign tags. so this'll be a minute. | 20:47 |
| SpamapS | \o/ | 20:52 |
| SpamapS | release day | 20:52 |
| corvus | okay both have been pushed | 20:52 |
| *** hashar has quit IRC | 20:53 | |
| corvus | SpamapS: fyi we're trying to keep the github changes to a mininum in this release, and then in the next, we're planning on having pabelanger and tobiash's work on retries, additional efficency, and (some) parallelization | 20:54 |
| corvus | (master is already slightly ahead of both of these releases) | 20:55 |
| SpamapS | corvus:cool. We don't have any problems with github at the moment, but we are about to migrate from github.com -> internal GHE | 20:56 |
| SpamapS | Meanwhile.. one struggle I'm really having is files matchers and the monorepo approach. | 21:02 |
| SpamapS | Using files matchers basically renders zuul enqueue/enqueue-ref unusable. | 21:03 |
| SpamapS | even if all I do is say irrelevant-files: '.*\.md$' | 21:04 |
| SpamapS | that matcher will never match on a CLI enqueue | 21:04 |
| corvus | SpamapS: i may have something for you, lemme dig it up | 21:04 |
| SpamapS | I wonder if we can figure something else out | 21:04 |
| SpamapS | Right now al I can do is go dig out the github webhook event and resend it. | 21:05 |
| corvus | SpamapS: https://review.opendev.org/660856 | 21:06 |
| corvus | SpamapS: making sure that case is handled is more or less the reason for my -1 on that. but i think it can be handled in a fairly straightforward way, so i expect we can merge a revision of that soon | 21:07 |
| corvus | SpamapS: (there is a workaround described in the first comment there -- which, briefly, is "don't add file matchers to job definitions, only add them when attaching jobs to pipelines". it's not as elegant as fixing zuul to match our mental models, but that's something you might consider doing (depending on how complex that would be) until we can revise and merge that change. | 21:09 |
| SpamapS | I only have files matchers on the project attachments. :-/ | 21:10 |
| SpamapS | Because I attach this job in 2 projects. | 21:10 |
| SpamapS | Still can't enqueue it. | 21:11 |
| corvus | SpamapS: oops, i may not understand then | 21:11 |
| SpamapS | The problem is that the files list comes from the trigger. | 21:11 |
| SpamapS | and the files list is empty when you trigger with the zuul CLI | 21:11 |
| corvus | SpamapS: are you enqueing a change or a ref? | 21:12 |
| SpamapS | change | 21:12 |
| SpamapS | also before we go much further.. I just found a bug in my files matchers, so it's possible the change I enqueued missed for that.. and I'm cray.. and you can stand down. | 21:12 |
| SpamapS | corvus: ok ya, files matcher was missing an entry. So that might be leftover PTSD from when I was using refs. | 21:15 |
| corvus | SpamapS: i *think* that Change objects which arrive to be enqueued that have incomplete file information are supposed to cause the mergers to go fetch their files, so in principle that case should be handled. but enqueing from the cli is slightly different than normal triggers, so there may be an oversight there.... | 21:15 |
| corvus | SpamapS: \o/ | 21:15 |
| corvus | SpamapS: at any rate, let's proceed under the assumption that (once the obvious error is corrected) your setup should work as you expect | 21:16 |
| SpamapS | corvus: yep, it's winding through the tubes now | 21:16 |
| corvus | and if it doesn't, there's a bug (but not a mis-design) | 21:16 |
| SpamapS | I really wish there was some way to test promote jobs more thorougly locally. I find myself iterating a lot by pushing PR's into master and bugging folks for approvals whenever I'm in this situation. | 21:17 |
| SpamapS | Have landed like 5 "oops forgot a -" and "need to widen file matcher" commits today. :-P | 21:17 |
| corvus | SpamapS: yeah, i tried a thing recently where i stuck a promote job in the gate pipeline so it would self-test, but there's a nuance with how zuul caches artifact queries that renders that dangerous, so i wouldn't recommend it yet (we might be able to make zuul more robust in that situation) | 21:18 |
| SpamapS | corvus:yeah, in my case, it's the zuul config itself that I keep getting slightly wrong. vars, matchers, etc. | 21:19 |
| SpamapS | I like the idea of having a gate job for testing it, but stops short of mutating anything. | 21:20 |
| SpamapS | Like "if we WERE going to deploy this... would it work up to the actual deploy poart?" | 21:20 |
| *** rfolco is now known as rfolco_off | 21:26 | |
| corvus | releases are published and announcement emails sent | 21:30 |
| SpamapS | woot | 21:33 |
| *** jamesmcarthur has joined #zuul | 21:40 | |
| *** jamesmcarthur has quit IRC | 21:51 | |
| *** sgw has quit IRC | 22:17 | |
| *** zbr|ruck has quit IRC | 22:17 | |
| *** rlandy is now known as rlandy|bbl | 22:24 | |
| openstackgerrit | James E. Blair proposed zuul/nodepool master: WIP: new devstack-based functional job https://review.opendev.org/665023 | 22:29 |
| *** sgw has joined #zuul | 22:37 | |
| *** rfolco_off has quit IRC | 22:56 | |
| *** dmsimard has quit IRC | 23:08 | |
| *** dmsimard0 has joined #zuul | 23:08 | |
| SpamapS | hm, confusing log messages today.. | 23:30 |
| SpamapS | http://paste.openstack.org/show/753206/ | 23:30 |
| SpamapS | So I have a project, GoodMoney/funnel-cake, and it has a zuul config. But zuul never runs any jobs, not even noops, on it. | 23:31 |
| SpamapS | "not in pipeline check" | 23:31 |
| SpamapS | derp | 23:32 |
| SpamapS | just figured it out | 23:32 |
| SpamapS | no branch protection | 23:32 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!