| openstackgerrit | Merged zuul/zuul-jobs master: test-upload-logs-swift: add download script as an artifact https://review.opendev.org/713998 | 00:01 |
|---|---|---|
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: [dnm] test unittests with base-test https://review.opendev.org/714006 | 00:07 |
| *** mattw4 has quit IRC | 00:11 | |
| *** jamesmcarthur has quit IRC | 00:17 | |
| *** jamesmcarthur has joined #zuul | 00:18 | |
| *** jamesmcarthur has quit IRC | 00:19 | |
| *** jamesmcarthur has joined #zuul | 00:19 | |
| openstackgerrit | Merged zuul/zuul-jobs master: Replace occurences of '/' in nodename to something more path friendly https://review.opendev.org/713182 | 00:21 |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: upload-logs-swift: Create a download script https://review.opendev.org/592341 | 00:25 |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: upload-logs-swift: Add a unicode file https://review.opendev.org/592853 | 00:25 |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: Revert "upload-logs-swift: test bulk download script change" https://review.opendev.org/714000 | 00:25 |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: test-upload-logs-swift: move script artifact after upload https://review.opendev.org/714007 | 00:25 |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: [dnm] test unittests with base-test https://review.opendev.org/714006 | 00:31 |
| *** jamesmcarthur has quit IRC | 00:31 | |
| *** jamesmcarthur has joined #zuul | 00:32 | |
| *** jamesmcarthur has quit IRC | 00:37 | |
| *** tosky has quit IRC | 00:54 | |
| *** jamesmcarthur has joined #zuul | 01:01 | |
| openstackgerrit | Merged zuul/zuul-jobs master: test-upload-logs-swift: move script artifact after upload https://review.opendev.org/714007 | 01:36 |
| *** rlandy has quit IRC | 01:41 | |
| *** rlandy has joined #zuul | 02:03 | |
| *** swest has quit IRC | 02:05 | |
| *** rlandy has quit IRC | 02:08 | |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: upload-logs-swift: Create a download script https://review.opendev.org/592341 | 02:16 |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: upload-logs-swift: Add a unicode file https://review.opendev.org/592853 | 02:16 |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: Revert "upload-logs-swift: test bulk download script change" https://review.opendev.org/714000 | 02:16 |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: test-upload-logs-swift: fix download link https://review.opendev.org/714016 | 02:16 |
| *** swest has joined #zuul | 02:20 | |
| *** jamesmcarthur has quit IRC | 02:27 | |
| *** jamesmcarthur has joined #zuul | 02:28 | |
| *** jamesmcarthur has quit IRC | 02:33 | |
| openstackgerrit | Merged zuul/zuul-jobs master: test-upload-logs-swift: fix download link https://review.opendev.org/714016 | 02:40 |
| *** jamesmcarthur has joined #zuul | 02:55 | |
| *** jamesmcarthur has quit IRC | 02:56 | |
| *** jamesmcarthur has joined #zuul | 02:56 | |
| *** swest has quit IRC | 03:01 | |
| *** jamesmcarthur has quit IRC | 03:19 | |
| *** jamesmcarthur has joined #zuul | 03:20 | |
| *** swest has joined #zuul | 03:21 | |
| *** jamesmcarthur has quit IRC | 03:25 | |
| *** SpamapS has quit IRC | 03:31 | |
| *** SpamapS has joined #zuul | 03:31 | |
| *** jamesmcarthur has joined #zuul | 03:43 | |
| *** jamesmcarthur has quit IRC | 03:43 | |
| *** jamesmcarthur has joined #zuul | 03:43 | |
| *** raukadah is now known as chandankumar | 04:24 | |
| *** bhavikdbavishi has joined #zuul | 04:44 | |
| *** klindgren_ has joined #zuul | 05:20 | |
| *** klindgren has quit IRC | 05:20 | |
| *** evrardjp has quit IRC | 05:36 | |
| *** evrardjp has joined #zuul | 05:36 | |
| AJaeger | ianw: I'm fine with the download script and gave a +2 on https://review.opendev.org/#/c/592341 | 05:37 |
| ianw | AJaeger: thanks! it's been through quite some iterations!! | 05:38 |
| AJaeger | ianw: yeah, noticed that you've been busy while I was asleep ;) | 05:42 |
| *** jamesmcarthur has quit IRC | 05:47 | |
| *** jamesmcarthur has joined #zuul | 06:15 | |
| *** jamesmcarthur has quit IRC | 06:44 | |
| *** jamesmcarthur has joined #zuul | 06:45 | |
| *** jamesmcarthur has quit IRC | 06:46 | |
| *** dpawlik has joined #zuul | 07:21 | |
| *** jamesmcarthur has joined #zuul | 07:26 | |
| *** jamesmcarthur has quit IRC | 07:44 | |
| *** jpena|off is now known as jpena | 08:12 | |
| *** jcapitao has joined #zuul | 08:40 | |
| *** tosky has joined #zuul | 08:51 | |
| *** harrymichal has joined #zuul | 09:17 | |
| *** harrymichal has quit IRC | 09:22 | |
| *** hashar has joined #zuul | 10:08 | |
| *** harrymichal has joined #zuul | 10:23 | |
| *** arxcruz|rover is now known as arxcruz|off | 10:56 | |
| *** hashar has quit IRC | 11:24 | |
| zbr | how can we define some tox jobs where sudo access is not removed? I have a use case for this. | 11:48 |
| *** panda has joined #zuul | 11:48 | |
| *** dpawlik has quit IRC | 11:50 | |
| AJaeger | what's the use case? | 11:59 |
| zbr | we have some molecule jobs using delegated mode, which are supposed to use the current nodeset for testing, which obviously needs root access. | 12:02 |
| mordred | zbr: the run playbook for the tox job is "roles: ['tox','revoke-sudo'] - you can make a child job that doesn't revoke sudo by supplying a run playbook - just make sure to put the tox role in it :) | 12:02 |
| zbr | ok, panda already proposed such a change two mins ago, hopefully it will work. (that is for rdo use) | 12:03 |
| mordred | cool | 12:03 |
| zbr | i know is it a bit counter-intuitive, as tox was designed not to mess with the system | 12:03 |
| mordred | yeah - but we use tox for things that arent' always tox-like :) | 12:04 |
| zbr | but we do have a fail-safe in place: the default settings are to skip delegated tests, so you need to mention extra params to run them. | 12:04 |
| zbr | my concern for local dev envs is a prio, i do not like surprises from tox. | 12:04 |
| zbr | did any of you had experience with nox? if so, of what nature? | 12:05 |
| mordred | zbr: haven't used it - although I tend to not like things that use a programming language as a config file format | 12:07 |
| mordred | (there's a reason we wrote pbr and don't use python in setup.py) | 12:08 |
| zbr | and python packaging tool decided that pbr sucks and they can do it better with TOML? ;) | 12:09 |
| zbr | i like pbr, proved to be far more reliable than setuptools. | 12:09 |
| mordred | \o/ | 12:09 |
| mordred | zbr: I'm not a fan of toml in general, I really don't like a thing that's named after its author and written "because I don't want to read the yaml docs" | 12:10 |
| mordred | so I tend to have a low esteem for choices to use it ;) | 12:10 |
| zbr | well, you are not alone here, to me it was more of a NIH syndrome example. | 12:11 |
| mordred | yup | 12:11 |
| panda | let's rename linux to openunix | 12:11 |
| zbr | instead of working with others to fix the problems of yaml, he decided he can do better. | 12:11 |
| mordred | I mean - I'm still old enough that I think of ini format as "the windows config format" - so extending it just seems extra gross | 12:11 |
| mordred | yeah | 12:11 |
| *** jcapitao is now known as jcapitao_lunch | 12:13 | |
| panda | zbr: it worked. | 12:13 |
| panda | zbr: I have a change before merging | 12:13 |
| panda | zbr: PS2 ready. | 12:13 |
| mordred | woot | 12:13 |
| panda | mordred: :) | 12:14 |
| avass | zbr: mordred: I have a change that makes it possible to toggle wether to revoke sudo https://review.opendev.org/#/c/706248/ | 12:19 |
| *** dpawlik has joined #zuul | 12:21 | |
| zbr | avass: nice start, but i would suggest reusing var name. lets see what others say about it. | 12:25 |
| avass | zbr: Yeah that makes sense | 12:26 |
| zbr | but wait before changing, lets see what mordred things | 12:27 |
| avass | sure | 12:27 |
| avass | I think I wanted to make sure it was separate so you don't accidentally revoke sudo for something you don't want to | 12:28 |
| zbr | in fact the condition could be included inside the role to avoid repeating the when condition | 12:29 |
| avass | I mean keeping sudo for a job where you want to revoke it | 12:30 |
| mordred | yeah - I was about to make the same suggestion | 12:30 |
| mordred | avass: if you put the when inside of revoke sudo itself ... | 12:30 |
| mordred | then name the variable revoke_sudo or something similar | 12:31 |
| avass | got it :) | 12:31 |
| mordred | avass: you could even get clever then I think - and maybe add a variable to your static nodes and then set revoke-sudo based on that variable ... I think there's a way to set things on nodepool nodes that show up in zuul variables? | 12:33 |
| mordred | (because it seems like as an admin it would be desirable in this case to be able to say "when you're running on this node, please do not run revoke-sudo") | 12:33 |
| avass | I think setting variable i nodepool is something we wanted but doesn't exist | 12:34 |
| avass | setting variable on the pipeline level would have been nice too | 12:35 |
| mordred | I could have sworn we added that | 12:36 |
| mordred | Shrews: am I just losing my mind here? | 12:36 |
| mordred | avass: I agree - I don't see any mention of setting arbitrary inventory variables in nodepool's config. I must just be hallucinating | 12:37 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Adds variable to toggle whether to revoke sudo https://review.opendev.org/706248 | 12:39 |
| avass | ho about that instead | 12:39 |
| avass | mordred: I think what we wanted it for was to know which drive to work on. 'C:' 'D:' etc but we've solved that in other ways instead | 12:40 |
| avass | and windows 7 machines has been a lot of pain | 12:41 |
| mordred | I'm sure they have been - we had a windows 7 build server for Drizzle back in the day - but it wasn't even connected to hudson because it was too much of a pita - the way it worked was it sad under my desk and I'd occasionally connect and run a build by hand | 12:43 |
| avass | wow | 12:44 |
| mordred | zbr: check out avass' patch: https://review.opendev.org/#/c/706262/ for making sudo more general. it seems like it's got a more complete version of one of the tasks in your https://review.opendev.org/#/c/703065 | 12:46 |
| mordred | of course, the sudo toggle patch above conflicts with the general patch - so that'll need to be sorted | 12:49 |
| avass | shouldn't be too hard to fix | 12:50 |
| mordred | nah. should be easy :) | 12:53 |
| mordred | might want to move main.yaml to be called revoke.yaml and then make a main.yaml that does include_tasks: revoke.yaml when: revoke_sudo - just cause there's more tasks in the expanded version and adding revoke_sudo to each of them might get boring | 12:54 |
| Shrews | mordred: i vote for the losing your mind option | 12:54 |
| mordred | ok | 12:54 |
| mordred | *phew* | 12:54 |
| Shrews | but also let me read sb a bit more (just returned from morning run) | 12:55 |
| mordred | Shrews: wow. that's so athletic | 12:56 |
| avass | mordred: ah yeah, was looking for a meta: end_role to solve that but couldn't find it. i guess include_tasks should work | 12:56 |
| mordred | I walked downstairs this morning - does that count as exercise? | 12:56 |
| Shrews | well... lack of activity due to virus things sucks and forces such things | 12:56 |
| mordred | avass: we use that pattern in a few places - it also has the nice effect of not listing each task in the included file and saying "skipped" -so it's a nice way to skip a block of tasks cleanly in the log | 12:57 |
| mordred | (logs much nicer than putting a set of tasks in a block and then putting a when on the block) | 12:57 |
| avass | I'll update it | 12:57 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Adds variable to toggle whether to revoke sudo https://review.opendev.org/706248 | 13:00 |
| *** jcapitao_lunch is now known as jcapitao | 13:03 | |
| Shrews | mordred: oh, there is this: https://zuul-ci.org/docs/nodepool/configuration.html#attr-providers.[openstack].pools.node-attributes | 13:04 |
| mordred | Shrews, avass: so that ^^ should allow setting a key-value pair and having that show up in the nodepool. dict on the node in the zuul inventory, yeah? | 13:05 |
| Shrews | i'm not certain how zuul manages that data, tbh. i'd have to re-review | 13:05 |
| mordred | nod | 13:06 |
| Shrews | shower then coffee taking precedence atm | 13:06 |
| mordred | well - if it does get passed through - perhaps we could define a key, say "revoke-sudo" that someone could set on a node doing that - and then in the revoke_sudo role have the default value of revoke_sudo be {{ nodepool.attributes['revoke-sudo'] | default(true) }} | 13:07 |
| avass | mordred, shrews: I'm going to guess it should be possible to extend it for the other drivers as well | 13:07 |
| mordred | and then docuemnt that behavior | 13:07 |
| avass | yeah that sounds good | 13:07 |
| mordred | probably want to get other people to weigh in on that as a possible approach before going _too_ far along ... it might be a bad idea for some reason I can't think of right now - I am still on my first coffee | 13:08 |
| avass | probably going create a role to upload logs to s3, might push that zuul-jobs if I get it working good enough | 13:17 |
| *** klindgren_ has quit IRC | 13:17 | |
| *** klindgren has joined #zuul | 13:17 | |
| *** jpena is now known as jpena|lunch | 13:23 | |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Adds variable to toggle whether to revoke sudo https://review.opendev.org/706248 | 13:31 |
| mordred | avass: yes please. hopefully between upload-logs-swift and upload-logs-gce there's enough to cargo-cult from to get you started | 13:32 |
| avass | mordred: was thinking of doing something like that :) | 13:36 |
| mordred | :) | 13:36 |
| -openstackstatus- NOTICE: Gerrit (review.opendev.org) will be down for maintenance starting at 14:00 (in less than 20 mins), probably until 15:00 UTC | 13:44 | |
| *** bhavikdbavishi has quit IRC | 13:45 | |
| openstackgerrit | Merged zuul/zuul master: Fix minor spelling error https://review.opendev.org/711926 | 13:46 |
| avass | what's the difference beteween zuul-executor pause and stop? | 13:56 |
| corvus | avass: stop is immediate and aborts jobs; pause stops running new jobs but lets existing ones continue | 13:57 |
| avass | ah | 13:57 |
| avass | so, feels like something similar should exist for mergers | 13:58 |
| *** sugaar has quit IRC | 14:03 | |
| *** jpena|lunch is now known as jpena | 14:05 | |
| *** klindgren has quit IRC | 14:10 | |
| -openstackstatus- NOTICE: The Gerrit service on review.opendev.org is offline for maintenance until 15:00 UTC http://lists.opendev.org/pipermail/service-announce/2020-March/000001.html | 14:10 | |
| *** ChanServ changes topic to "The Gerrit service on review.opendev.org is offline for maintenance until 15:00 UTC http://lists.opendev.org/pipermail/service-announce/2020-March/000001.html" | 14:10 | |
| *** klindgren has joined #zuul | 14:10 | |
| *** harrymichal has quit IRC | 14:46 | |
| mordred | avass: tobiash agrees and has a patch up for that | 14:47 |
| *** harrymichal has joined #zuul | 14:50 | |
| *** sugaar has joined #zuul | 14:54 | |
| avass | mordred: nice | 14:55 |
| *** ChanServ changes topic to "Discussion of the project gating system Zuul | Website: https://zuul-ci.org/ | Docs: https://zuul-ci.org/docs/ | Source: https://git.zuul-ci.org/ | Channel logs: http://eavesdrop.openstack.org/irclogs/%23zuul/ | Weekly updates: https://etherpad.openstack.org/p/zuul-update-email" | 15:07 | |
| -openstackstatus- NOTICE: Gerrit maintenance is concluded at this time and requested renames have been performed | 15:07 | |
| *** hashar has joined #zuul | 15:09 | |
| mordred | avass: https://review.opendev.org/#/c/707192/ | 15:20 |
| mordred | avass: already merged even | 15:20 |
| *** toabctl has quit IRC | 15:31 | |
| tobiash | avass: might not be in the current release yet though | 15:36 |
| *** zxiiro has joined #zuul | 15:38 | |
| *** mattw4 has joined #zuul | 15:43 | |
| *** Tahvok has joined #zuul | 15:44 | |
| *** jcapitao is now known as jcapitao_afk | 15:50 | |
| *** jamesmcarthur has joined #zuul | 15:54 | |
| *** mattw4 has quit IRC | 15:59 | |
| *** mattw4 has joined #zuul | 16:00 | |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-operator master: Add utility files and scripts https://review.opendev.org/714163 | 16:14 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-operator master: Fix executor private key permission issue https://review.opendev.org/714164 | 16:14 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-operator master: Add integration test playbook https://review.opendev.org/714165 | 16:14 |
| *** jcapitao_afk is now known as jcapitao | 16:24 | |
| tobiash | zuul-maint: I just found yet another race that leads to a node leak | 16:30 |
| corvus | tobiash: you are good at that | 16:31 |
| tobiash | from what I see in the logs the sequence is 1. node request fulfilled 2. Canceling node request 3. Accepting node request | 16:32 |
| *** hashar has quit IRC | 16:32 | |
| *** hashar has joined #zuul | 16:32 | |
| tobiash | I didn't dig into the code yet, I'll do a further analysis next week probably | 16:33 |
| *** hashar has quit IRC | 16:35 | |
| *** hashar has joined #zuul | 16:35 | |
| corvus | tobiash: that sounds like a sequence we anticipated at least | 16:37 |
| *** andreykurilin has joined #zuul | 16:40 | |
| tobiash | seems like it didn't ran into the request.canceled if clause | 16:40 |
| *** openstack has joined #zuul | 16:58 | |
| *** ChanServ sets mode: +o openstack | 16:58 | |
| corvus | andreykurilin: if you have access to the decrypted version of the secret, you can define the same data as a variable locally | 16:59 |
| corvus | but yes, if you don't have access to that, that's not possible | 16:59 |
| corvus | (though, depending on what the service is, maybe you could create a test account or something) | 16:59 |
| andreykurilin | ok... sounds like a task for 10 minutes email | 17:00 |
| andreykurilin | :) | 17:00 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-operator master: Add integration test playbook https://review.opendev.org/714165 | 17:04 |
| AJaeger | zbr, btw. we have openstack-tox-with-sudo - isn't that what you need? | 17:10 |
| *** toabctl has joined #zuul | 17:19 | |
| zbr | AJaeger: i am afraid that this may not be of much help for all cases, mostly because it prevents inheriting from other tox jobs. | 17:20 |
| zbr | i kinda like the idea to control sudo via a variable, like avass proposed earlier. | 17:20 |
| zbr | now I do not know if there are other problems with variable approach, like security concerns. clearly is a good idea to have revoke-sudo as default behavior. | 17:21 |
| zbr | in fact on ansible zuul instance i often need sudo for tools/test-setup.sh part, but later sudo is revoked, before tox is invoked. | 17:22 |
| zbr | this allowed me to do a git of extra tunning needed, like starting a service. | 17:23 |
| AJaeger | I see... | 17:24 |
| *** jcapitao has quit IRC | 17:24 | |
| fungi | i don't think we rely on revoke-sudo for security controls, just to provide a consistent way for jobs to provide feedback when a payload unexpectedly tries to do something as root | 17:29 |
| fungi | to prevent calls to sudo from slipping past review and ending up merged in repositories | 17:30 |
| *** avass has quit IRC | 17:34 | |
| *** evrardjp has quit IRC | 17:36 | |
| *** evrardjp has joined #zuul | 17:36 | |
| *** jamesmcarthur has quit IRC | 17:37 | |
| *** jpena is now known as jpena|off | 17:44 | |
| *** jamesmcarthur has joined #zuul | 17:49 | |
| *** harrymichal has quit IRC | 18:04 | |
| *** harrymichal has joined #zuul | 18:04 | |
| *** hashar has quit IRC | 18:08 | |
| *** hashar has joined #zuul | 18:09 | |
| *** hashar has quit IRC | 18:18 | |
| *** hashar has joined #zuul | 18:18 | |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-operator master: Add integration test playbook https://review.opendev.org/714165 | 18:20 |
| openstackgerrit | Jimmy McArthur proposed zuul/zuul-website master: Promoting Zuul User Survey https://review.opendev.org/714188 | 18:45 |
| *** rlandy has joined #zuul | 18:46 | |
| *** jamesmcarthur has quit IRC | 18:47 | |
| *** hashar is now known as hasharAway | 18:49 | |
| *** bhavikdbavishi has quit IRC | 18:51 | |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-operator master: Add integration test playbook https://review.opendev.org/714165 | 18:52 |
| *** jamesmcarthur has joined #zuul | 19:09 | |
| *** hasharAway has quit IRC | 19:10 | |
| *** bhavikdbavishi has joined #zuul | 19:21 | |
| *** nhicher has quit IRC | 19:29 | |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-operator master: Add integration test playbook https://review.opendev.org/714165 | 19:34 |
| *** jamesmcarthur has quit IRC | 19:44 | |
| *** jamesmcarthur has joined #zuul | 19:45 | |
| *** jamesmcarthur has quit IRC | 19:45 | |
| *** jamesmcarthur has joined #zuul | 19:49 | |
| openstackgerrit | Jimmy McArthur proposed zuul/zuul-website master: Promoting Zuul User Survey https://review.opendev.org/714188 | 19:55 |
| *** nhicher has joined #zuul | 19:55 | |
| openstackgerrit | Jimmy McArthur proposed zuul/zuul-website master: Promoting Zuul User Survey https://review.opendev.org/714188 | 20:00 |
| tristanC | corvus: the zuul-operator functional test finally worked, here is the builds api output of a SUCCESS job: https://zuul.opendev.org/t/zuul/build/2ef909ed0949401bb7aef9dc76225c09/log/job-output.txt#1001-1021 | 20:08 |
| *** harrymichal has quit IRC | 20:14 | |
| *** harrymichal has joined #zuul | 20:15 | |
| tristanC | next i'll look into adding a console stream check | 20:19 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: use-buildset-registry: protect against /var/user/1000 not existing https://review.opendev.org/714216 | 20:28 |
| corvus | tristanC: \o/ sorry, i'm deep in opendev ops today, i look forward to rejoining you on that next week :) | 20:29 |
| *** jamesmcarthur has quit IRC | 21:16 | |
| *** jamesmcarthur has joined #zuul | 21:17 | |
| *** harrymichal has quit IRC | 21:19 | |
| *** harrymichal has joined #zuul | 21:20 | |
| *** jamesmcarthur has quit IRC | 21:22 | |
| *** harrymichal has quit IRC | 21:25 | |
| *** harrymichal has joined #zuul | 21:25 | |
| *** harrymichal has quit IRC | 21:35 | |
| *** harrymichal has joined #zuul | 21:35 | |
| mordred | tristanC: if you have a spare sec, https://review.opendev.org/#/c/714216/ could do with a +A - we verified it in a depends-on with ps 8 of https://review.opendev.org/#/c/714171/8 | 21:40 |
| *** harrymichal has quit IRC | 21:40 | |
| mordred | fungi, Shrews: ^^ | 21:40 |
| *** harrymichal has joined #zuul | 21:40 | |
| *** jamesmcarthur has joined #zuul | 21:47 | |
| *** rlandy has quit IRC | 21:48 | |
| *** jamesmcarthur has quit IRC | 21:53 | |
| *** jamesmcarthur has joined #zuul | 22:07 | |
| *** y2kenny has joined #zuul | 22:15 | |
| y2kenny | Hi, I have a question about job variables. if I defined one (say, job.var.foo="test message") how do I reference it in the ansible playbook? (let say if I have a debug task and want to print that var in a debug message) | 22:20 |
| openstackgerrit | Merged zuul/zuul-jobs master: use-buildset-registry: protect against /var/user/1000 not existing https://review.opendev.org/714216 | 22:21 |
| mordred | y2kenny: just as foo - so debug: "{{ foo }}" should totally just work | 22:26 |
| y2kenny | ok | 22:27 |
| mordred | y2kenny: the job variables are passed to the job as inventory variables | 22:27 |
| mordred | s/job/playbook | 22:27 |
| y2kenny | there's no additional namespacing of variables, etc? | 22:27 |
| mordred | nope. just as they're defined | 22:27 |
| y2kenny | cool. | 22:27 |
| y2kenny | another quick question. Do you have any recommendation on the best way to play with these zuul features? | 22:27 |
| y2kenny | the jobs and config being checked into git is great for production | 22:28 |
| mordred | it's all dynamic - so you can just propose a change with something new and it'll run that way before you commit it | 22:28 |
| y2kenny | but it's kind of slow when I am still learning. | 22:28 |
| mordred | s/commit/merge/ | 22:28 |
| mordred | that's mostly how I hack on things. :) | 22:29 |
| y2kenny | I guess I should just deploy the docker tutorial locally with compose | 22:29 |
| fungi | yeah, propose changes to jobs (or propose entire jobs) which are not in a trusted config project and you'll get speculative execution of the proposed additions/changes | 22:29 |
| y2kenny | ok. understood. | 22:29 |
| fungi | but sure, the quickstart environment is great for that too | 22:30 |
| mordred | yeah | 22:31 |
| fungi | the zuul quickstart is especially useful for testing things with trusted config repos and changes to pipelines which aren't triggered on proposed changes | 22:39 |
| *** jamesmcarthur has quit IRC | 22:41 | |
| *** jamesmcarthur has joined #zuul | 22:42 | |
| y2kenny | fungi: understood. I have been decomposing the quickstart to turn it into a production setup (so far have migrated zk into my own zk install on a k8s cluster... still need to try doing nodepool and executor on the k8s cluster) I probably should keep a separate plain quick start around just to play around with various job config | 22:44 |
| fungi | y2kenny: that vector of experimentation is very much appreciated! thanks for testing it out | 22:47 |
| y2kenny | another question (unrelated): how are job "fan out" handled? For example, I have a project that needs to be build for 3 different linux distro. They are independent and can be done in parallel. Does each of those distro needs to have its own build job? | 22:59 |
| mnaser | y2kenny: in that case, i would define a base job called my-project with all the necessary playbooks, then i would create 3 other jobs called my-project-ubuntu-bionic, my-project-centos-8, etc. and in those jobs, you would do "parent: my-project" and nothing else (hopefully) | 23:00 |
| mnaser | then inside your project definition, you would add those "os speciifc" jobs, that way, you dont have to maintain 3 different iterations of the same job | 23:01 |
| mnaser | (ansible constructs in this case like "package" which install things agnostically come in handy) | 23:01 |
| y2kenny | ok. That sounds good for this case. But what about test jobs that I may have hundreds that can be run in parallel and developer can add them very frequently. Does each of those test need to be defined in zuul individually or is there a way to facilitate this? | 23:05 |
| mordred | we've got a few thousand ourselves - yeah- just define them in some combination of central "job library" repos and then whenever/whereever the dev needs to add them. jobs definitions are cheap in zuul - we make and delete them all the time | 23:08 |
| mordred | y2kenny: as an example, if you look here: http://zuul.opendev.org/t/openstack/jobs you can scroll for a while and see the mass of jobs in the openstack tenant :) | 23:08 |
| y2kenny | cool | 23:09 |
| fungi | though it's worth pointing out that if you *do* need a job with nodes from three different distro images, zuul can totally do that too | 23:13 |
| fungi | just define a nodeset with those and set it on the job | 23:13 |
| fungi | and then reference the nodes by name in your playbooks | 23:14 |
| y2kenny | I have yet to try that and I will definitely need to do that | 23:14 |
| y2kenny | so hosts:nodesname instead of hosts: all? | 23:15 |
| mordred | yup | 23:15 |
| mordred | I mean - unless you have tasks that you want to run on all of the hosts | 23:15 |
| mordred | but yeh | 23:15 |
| y2kenny | um... so I am a bit confuse... doesn't the nodeset filter the nodes that a job will be executed on? (I haven't read up on the docs yet so I may have a wrong understanding) | 23:17 |
| *** harrymichal has quit IRC | 23:18 | |
| y2kenny | like... wouldn't the playbook's "all" just mean all the node in the nodeset? or is it typical to have a single playbook share by all job? | 23:18 |
| mordred | the nodeset defines what nodes you need for the job. it'll then either create them or check them out (depending on cloud vs static) | 23:18 |
| mordred | and then yes - it puts them into an inventory - so to the playbook "all" is all of the nodes defined in the nodeset for the job | 23:19 |
| mordred | we frequently use "all" in playbooks for simple job definitions just so we don't have to worry about node names since it really doens't matter - many times "all" of the one node the job needs is easy enough | 23:20 |
| *** armstrongs has joined #zuul | 23:20 | |
| mordred | for more complex multi-node jobs where the nodes actually have real differences that matter, definitely using names and refering to things purpusefully by name is the way to go | 23:20 |
| * mordred may not be answering the actual question ... | 23:20 | |
| y2kenny | I see. Make sense | 23:21 |
| fungi | yeah, if you're familiar with ansible, basically the hosts in your inventory are "localhost" (the zuul executor) and also any nodes in the nodeset attached to the job | 23:21 |
| mordred | yes - although in our case "all" does not match localhost | 23:21 |
| *** tjgresha_ has joined #zuul | 23:23 | |
| *** tjgresha has quit IRC | 23:23 | |
| *** jamesmcarthur has quit IRC | 23:24 | |
| y2kenny | ok. Another unrelated question. I remember you guys mentioning zuul can handle a restarting node. Do you have an example for that (or the relevant terms that I should dig into?) I assume the executor is the one in charge of pinging the restarting node and potentially issue a fail if there's a timeout? | 23:25 |
| *** portdirect has quit IRC | 23:25 | |
| *** evrardjp has quit IRC | 23:25 | |
| fungi | the only gotcha i know about is if you're relying on console logging you need to restart the console log daemon after a node reboot | 23:26 |
| *** evrardjp has joined #zuul | 23:27 | |
| *** portdirect has joined #zuul | 23:27 | |
| fungi | and i think we have an ansible role you can add after the reboot for exactly that... checking now | 23:27 |
| *** irclogbot_0 has quit IRC | 23:27 | |
| *** armstrongs has quit IRC | 23:28 | |
| *** irclogbot_3 has joined #zuul | 23:29 | |
| *** Defolos has quit IRC | 23:31 | |
| y2kenny | ok thanks | 23:32 |
| *** Defolos has joined #zuul | 23:33 | |
| *** gcerami_ has joined #zuul | 23:35 | |
| *** guilhermesp_ has joined #zuul | 23:36 | |
| *** panda has quit IRC | 23:37 | |
| *** guilhermesp has quit IRC | 23:37 | |
| *** gcerami_ is now known as panda | 23:37 | |
| *** guilhermesp_ is now known as guilhermesp | 23:37 | |
| *** evrardjp has quit IRC | 23:38 | |
| fungi | y2kenny: i think you just add this role after the reboot(s) https://zuul-ci.org/docs/zuul-jobs/general-roles.html#role-start-zuul-console | 23:40 |
| mordred | yes. well - after the root and the wait_for | 23:40 |
| mordred | s/root/reboot | 23:41 |
| y2kenny | ok | 23:41 |
| *** evrardjp has joined #zuul | 23:43 | |
| * mordred is looking for a job example - but unfortunately "reboot" is a common word to find across the openstack code base :) | 23:43 | |
| fungi | right, i struggled to locate a clear example unfortunately | 23:44 |
| fungi | ianw is the last person i recall trying it in earnest | 23:44 |
| *** jamesmcarthur has joined #zuul | 23:44 | |
| mordred | yeah | 23:45 |
| *** jamesmcarthur has quit IRC | 23:50 | |
| tristanC | mordred: lgtm, though why not creating the dir (/var/user/1000) if it doesn't exists? | 23:51 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-operator master: Add integration test playbook https://review.opendev.org/714165 | 23:52 |
| mordred | tristanC: the commit message is a typo - it's /run/user/1000 - and I think it's cause /run isn't a thing on that old ubuntu | 23:54 |
| mordred | but the only things that need that aren't going to run on xenial really well anyway :) | 23:54 |
| *** jamesmcarthur has joined #zuul | 23:55 | |
| fungi | the /run ramdisk likely exists on xenial, but you need logind (and so probably bionic) to create the /run/user/* dirs | 23:55 |
| tristanC | oh i see. well either way, that sounds ok :) | 23:57 |
| *** tosky has quit IRC | 23:59 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!