| openstackgerrit | Merged openstack-infra/zuul master: Don't report non-live items in stats https://review.openstack.org/605540 | 00:09 |
|---|---|---|
| *** rlandy has quit IRC | 00:36 | |
| openstackgerrit | Ian Wienand proposed openstack-infra/nodepool master: Normalise more of the API stats calls https://review.openstack.org/605898 | 01:12 |
| *** dkehn has quit IRC | 01:24 | |
| *** dkehn has joined #zuul | 01:28 | |
| tristanC | Shrews: iirc, minikube needs a nested vm, we talked about using kubespray instead. | 01:40 |
| tristanC | Shrews: corvus: clarkb: having the nodepool deployment bits refactor would be very appreciated for the k8s integration job indeed | 01:41 |
| tristanC | corvus: atm, here is how zuul process the connection_port data: https://review.openstack.org/#/c/570668/9/zuul/executor/server.py | 01:42 |
| clarkb | tristanC: minikube has a mode to install on localhost without a VM | 01:44 |
| clarkb | --vm-driver=none | 01:44 |
| tristanC | clarkb: oh then that may works. the issue remains to sequence "setup provider (openstack/k8s)", "setup nodepool", "configure nodepool" and "run integration test", is it possible to have roles/playbook for each of these steps so that i only have to write step1 and step3 ? | 01:55 |
| clarkb | tristanC: I think the discussuon earlier was to make thr nodepool setup/config split out | 01:56 |
| clarkb | and then the integration test today is just loop and check if expected image and node are created | 01:56 |
| tristanC | what about zuul integration, shouldn't we also setup zuul and configure a git driver to test jobs too? | 01:58 |
| clarkb | we dont do that today so that would be new | 01:58 |
| *** dkehn has quit IRC | 02:04 | |
| clarkb | in the past the biggest concerns have been the integration between nodepool and its provisioner apis | 02:09 |
| clarkb | thr api between zuul and nodepool is something we control and have unittests for etc | 02:09 |
| tristanC | one of the issue to do such test is that it's not easy to trigger a job manually in zuul, e.g. we would have to setup a periodic pipeline and wait for execution | 02:12 |
| clarkb | you could use zuul-enqueue | 02:13 |
| tristanC | doesn't that needs a connection which supports trigger? | 02:13 |
| clarkb | maybe? I've never actually tried it with a non working connection | 02:14 |
| tristanC | having a zuul-base-job/zuul-jobs/zuul/nodepool integration test seems worhty imo, especially since kubectl connection relies on specific bits in each of those project | 02:14 |
| tristanC | well i mean that would be a nice addition :) | 02:15 |
| *** dkehn has joined #zuul | 03:39 | |
| tobiash | mnaser, Shrews, mordred, corvus: atm we don't do that, but it's on the backlog. In our old environment we changed min-ready by having two nodepool configs and a cronjob that sets symlinks to them. But actually this is a hack and not that easy in a kubernetes world. So I would love to see a way to express this in directly in the nodepool config. | 04:01 |
| *** toabctl has quit IRC | 04:48 | |
| *** quiquell|off is now known as quiquell | 05:41 | |
| *** chkumar|off is now known as chandankumar | 05:46 | |
| *** pcaruana has joined #zuul | 06:11 | |
| tristanC | tobiash: perhaps nodepool could anticipate resources usage based on history? | 06:12 |
| tobiash | tristanC: that would be the most generic and no config approach, but probably the most difficult way to implement | 06:14 |
| *** bhavikdbavishi has joined #zuul | 06:40 | |
| *** quiquell is now known as quiquell|brb | 07:04 | |
| *** ssbarnea|bkp has quit IRC | 07:16 | |
| *** quiquell|brb is now known as quiquell | 07:35 | |
| *** hashar has joined #zuul | 07:43 | |
| *** jpena|off is now known as jpena | 07:43 | |
| *** quiquell is now known as quiquell|brb | 07:55 | |
| *** quiquell|brb is now known as quiquell | 08:05 | |
| openstackgerrit | Ian Wienand proposed openstack-infra/nodepool master: Normalise more of the API stats calls https://review.openstack.org/605898 | 08:49 |
| *** toabctl has joined #zuul | 09:42 | |
| openstackgerrit | Merged openstack-infra/zuul master: replace dict.update by a dict merge in zuul_return https://review.openstack.org/602054 | 09:57 |
| quiquell | tristanC: Do you know where do we have the issue https://logs.rdoproject.org/70/16370/14/check/tox-docs/00152f6/job-output.txt.gz#_2018-09-26_21_11_27_059671 ? | 10:11 |
| quiquell | tristanC: returning empty data after reading zuul config here https://github.com/openstack-infra/zuul-sphinx/blob/master/zuul_sphinx/zuul.py#L107 | 10:12 |
| openstackgerrit | Ian Wienand proposed openstack-infra/zuul-sphinx master: Add attr-overview directive https://review.openstack.org/604980 | 10:18 |
| openstackgerrit | Ian Wienand proposed openstack-infra/nodepool master: Use zuul-sphinx for configuration layout https://review.openstack.org/604274 | 10:23 |
| openstackgerrit | Ian Wienand proposed openstack-infra/nodepool master: Add overview of config options https://review.openstack.org/604984 | 10:23 |
| *** bhavikdbavishi has quit IRC | 10:38 | |
| *** AJaeger has quit IRC | 10:56 | |
| *** rfolco has quit IRC | 11:07 | |
| *** jpena is now known as jpena|lunch | 11:07 | |
| *** AJaeger has joined #zuul | 11:08 | |
| *** ssbarnea|bkp has joined #zuul | 11:08 | |
| *** rfolco has joined #zuul | 11:26 | |
| *** ssbarnea|bkp has quit IRC | 11:40 | |
| *** panda|off is now known as panda | 11:42 | |
| *** yolanda has joined #zuul | 11:44 | |
| *** EmilienM is now known as EvilienM | 12:00 | |
| *** jpena|lunch is now known as jpena | 12:05 | |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: web: add tenant and project scoped, JWT-protected actions https://review.openstack.org/576907 | 12:10 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: CLI: add create-web-token command https://review.openstack.org/605386 | 12:10 |
| *** rfolco has quit IRC | 12:18 | |
| *** rlandy has joined #zuul | 12:24 | |
| openstackgerrit | neilsun proposed openstack-infra/zuul master: Add type check for zuul conf https://review.openstack.org/591917 | 12:25 |
| openstackgerrit | neilsun proposed openstack-infra/zuul master: Add type check for zuul conf https://review.openstack.org/591917 | 12:51 |
| *** hashar is now known as hasharAway | 12:51 | |
| *** bhavikdbavishi has joined #zuul | 12:55 | |
| openstackgerrit | Gabriele Cerami proposed openstack-infra/zuul-sphinx master: Raise an error if a file in zuul.d is empty https://review.openstack.org/606062 | 12:56 |
| *** tbarron has joined #zuul | 12:56 | |
| openstackgerrit | Gabriele Cerami proposed openstack-infra/zuul-sphinx master: Raise an error if a file in zuul.d is empty https://review.openstack.org/606062 | 12:58 |
| *** rfolco has joined #zuul | 13:07 | |
| tobiash | mordred: on fedora based openshift nodes the bionic version of bwrap crashes with 'setting up gid map: invalid argument' | 13:21 |
| tobiash | where the alpine based bwrap works... | 13:21 |
| mordred | tobiash: oh spectacular | 13:23 |
| *** samccann has joined #zuul | 13:23 | |
| mordred | tobiash: oh - and now I remember why I went with python:alpine vs. python:slim in pbrx - python:alpine has newer bubblewrap | 13:25 |
| tobiash | mordred: alpine: tried 0.2.0 and 0.3.0, both work... ubuntu bionic: uses 2.1.0 -> failed | 13:28 |
| *** bhavikdbavishi has quit IRC | 13:39 | |
| quiquell | Hello | 14:03 |
| quiquell | Have one question regarding variables at jobs | 14:04 |
| quiquell | Can i make cacheable facts inside zuul executions ? | 14:04 |
| quiquell | mordred: ^ you know ? | 14:04 |
| pabelanger | quiquell: zuul-executor is setup for cache facts today | 14:05 |
| tobiash | mordred: now tried 0.3.0 on bionic and it fails too | 14:05 |
| tobiash | mordred: so it's not the version | 14:06 |
| quiquell | pabelanger: It does not store the vars from jobs there ? Do I have to manually add them to the facts ? | 14:06 |
| quiquell | pabelanger: So I can use them in later ansibe-playbook executions ? | 14:06 |
| mordred | tobiash: I'm guessing it's an incompatibility with the libc in ubuntu and the kernel in fedora and the kernel capabilities stuff in bwrap | 14:07 |
| mordred | tobiash: which is ... wow | 14:07 |
| pabelanger | quiquell: do you mean set_facts? | 14:08 |
| mordred | quiquell: there is another mechanism you can use to pass a variable from one playbook to another | 14:08 |
| tobiash | mordred: so as I understood it so far it sets up a usernamespace and crashes when writing the gid_map | 14:08 |
| quiquell | mordred: We are dumping into a file and loading it, was thiking that maybe we can use cacheable facts so we don't need to dump | 14:08 |
| tobiash | mordred: writing the uid_map works | 14:08 |
| mordred | quiquell: https://zuul-ci.org/docs/zuul/user/jobs.html?highlight=zuul_return#return-values | 14:09 |
| mordred | tobiash: hrm | 14:09 |
| tobiash | mordred, quiquell: zuul_return only makes the variables available within a *child job*, not in a later playbook of the same job | 14:10 |
| pabelanger | set_fact: cacheable is a thing now, but you need to be explicity: https://docs.ansible.com/ansible/2.5/modules/set_fact_module.html | 14:10 |
| pabelanger | ansible 2.4+ | 14:10 |
| quiquell | tobiash, mordred: we are running playbooks inside jobs calling bash, so we lose context | 14:10 |
| mordred | tobiash: oh - silly me | 14:10 |
| tobiash | quiquell, mordred: there is a wip change that makes it possible however: https://review.openstack.org/549807 | 14:11 |
| mordred | quiquell: so this isn't zuul's ansible you're concerned with right now - it's ansible you're executing inside of yourjob content? | 14:11 |
| quiquell | mordred: yep... I want to read variables in the "vars" section ot our job inside the playbooks we run from bash | 14:12 |
| mordred | quiquell: so the issue you're running in to is that you have zuul job variables set, and then in your job you are executing ansible and you want those variables available to your in-job ansible | 14:12 |
| quiquell | mordred: We can always dump a file, but maybe there is more elegant solution like cacheable facts | 14:12 |
| quiquell | mordred: yep, with bash in the middle | 14:13 |
| quiquell | mordred: bash is the glue | 14:13 |
| mordred | gotit. so - fact caching isn't going to help - because the ansible zuul is running is executing in a protected context on the executor, and your ansible is running on the remote test node | 14:13 |
| mordred | I don't think, at least at the moment, that there is a better mechanism than the one you're using | 14:13 |
| quiquell | mordred: so dumping is the only solution | 14:13 |
| mordred | yah | 14:13 |
| quiquell | mordred: after we remove bash from the equation we will be able to use the vars directly ? | 14:14 |
| mordred | I've been pondering writing an ansible module for ansible so that you can have ansible execute ansible but with variables transferred and stuff | 14:14 |
| mordred | quiquell: if zuul is executing your playbooks, then yes | 14:14 |
| quiquell | mordred: ack | 14:17 |
| quiquell | thanks mordred, tobiash , pabelanger | 14:17 |
| *** quiquell is now known as quiquell|off | 14:31 | |
| *** electrofelix has quit IRC | 14:34 | |
| tbarron | why is zuul still running openstack-tox-py35 in e.g. https://review.openstack.org/#/c/605883/ (stable/queens) despite our having merged https://review.openstack.org/#/c/605893/ ? | 14:56 |
| corvus | answered in -infra | 15:03 |
| tobiash | mordred: I think I just f*** up my dockerfile for ubuntu... ubuntu's adduser defaults to nogroup while alpine's adduser defaults to the group with the same name as the user if it exists... | 15:16 |
| mordred | tobiash: AH! fingers crossed that fixing that fixes it | 15:20 |
| tbarron | corvus: ty! | 15:20 |
| *** bhavikdbavishi has joined #zuul | 15:23 | |
| tobiash | mordred: verified, that fixed it, so bwrap is not broken | 15:37 |
| gundalow | clarkb: hi. You mentioned it might be worth restarting etherpad (apache?) before Ansible's Contributors Summit, which starts on Monday | 15:43 |
| clarkb | gundalow: ya one of us can do that today to clear out any stale connections | 15:45 |
| gundalow | Thank you :) | 15:47 |
| Shrews | tristanC: what's wrong with needing a nested vm for minikube? devstack uses them too | 15:55 |
| clarkb | shrews thr actual containers run in that vm with minikube but we dont do anything except test ssh with devstack nested VMs | 15:56 |
| clarkb | but minikube runs on localhost too so non issue | 15:56 |
| Shrews | clarkb: i cannot find any reference to --vm-driver=None. https://kubernetes.io/docs/setup/minikube/#quickstart | 15:56 |
| clarkb | Shrews: https://github.com/kubernetes/minikube#quickstart | 15:58 |
| clarkb | has a list of the drivers | 15:58 |
| Shrews | ah | 15:58 |
| *** ianychoi has quit IRC | 16:39 | |
| *** ianychoi has joined #zuul | 16:44 | |
| *** jpena is now known as jpena|off | 16:57 | |
| openstackgerrit | Merged openstack-infra/zuul master: Fix node leak on job removal https://review.openstack.org/605527 | 17:35 |
| *** TheJulia is now known as needssleep | 18:02 | |
| *** hasharAway is now known as hasharRlyAwy | 18:22 | |
| *** hasharRlyAwy is now known as hasharAway | 19:42 | |
| *** openstackgerrit has quit IRC | 20:07 | |
| *** bhavikdbavishi has quit IRC | 20:12 | |
| *** rlandy is now known as rlandy|brb | 20:27 | |
| *** samccann has quit IRC | 20:40 | |
| *** rlandy|brb is now known as rlandy | 20:49 | |
| *** panda has quit IRC | 20:57 | |
| *** panda has joined #zuul | 20:58 | |
| *** rfolco has quit IRC | 21:18 | |
| *** hasharAway has quit IRC | 21:22 | |
| clarkb | gundalow: fyi I just restarted apache to claer out those stale connections if they are there | 21:53 |
| *** EvilienM is now known as EmilienM | 22:20 | |
| *** panda is now known as panda|off | 22:26 | |
| *** rlandy has quit IRC | 22:27 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!