| *** jamesmcarthur has joined #zuul | 00:20 | |
| *** jamesmcarthur has quit IRC | 00:26 | |
| *** tosky has quit IRC | 00:26 | |
| *** jamesmcarthur has joined #zuul | 00:58 | |
| *** jamesmcarthur has quit IRC | 01:17 | |
| *** spsurya has joined #zuul | 01:28 | |
| *** bhavikdbavishi has joined #zuul | 02:39 | |
| *** bhavikdbavishi has quit IRC | 02:41 | |
| *** bhavikdbavishi has joined #zuul | 02:42 | |
| *** bhavikdbavishi1 has joined #zuul | 02:44 | |
| *** bhavikdbavishi has quit IRC | 02:46 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 02:46 | |
| *** bstinson has quit IRC | 03:15 | |
| *** bstinson has joined #zuul | 03:22 | |
| *** bstinson has quit IRC | 03:39 | |
| *** bstinson has joined #zuul | 03:51 | |
| *** jamesmcarthur has joined #zuul | 05:46 | |
| *** jamesmcarthur has quit IRC | 05:50 | |
| *** raukadah is now known as chkumar|rover | 05:54 | |
| *** pcaruana has joined #zuul | 06:32 | |
| *** chkumar|rover is now known as chkumar|ruck | 06:51 | |
| *** jcapitao has joined #zuul | 08:01 | |
| *** jangutter has joined #zuul | 08:09 | |
| *** sugaar has joined #zuul | 08:24 | |
| *** tosky has joined #zuul | 08:31 | |
| *** hashar has joined #zuul | 08:46 | |
| *** jpena|off is now known as jpena | 08:48 | |
| *** ttx has quit IRC | 08:50 | |
| *** ttx has joined #zuul | 08:50 | |
| *** themroc has joined #zuul | 08:56 | |
| *** bogdando has joined #zuul | 10:02 | |
| bogdando | hi, could you please take a look https://review.opendev.org/#/c/696337/ ? | 10:02 |
|---|---|---|
| bogdando | I cannot apply it to my local zuul via depends-on for some reason... And I need this to use custom pypi mirror for fedora 31 subnodes | 10:03 |
| *** bogdando has left #zuul | 10:18 | |
| *** tosky_ has joined #zuul | 10:26 | |
| sugaar | Hi, somebody knows where nodepool takes the nodes from in the zuul/doc/source.admin/example demo?? | 10:26 |
| sugaar | I am trying to configure it to use some nodes in a kubernetes cluster in minikube but It can't find the .kub/config file. So I was wondering how it does it by default | 10:27 |
| *** tosky has quit IRC | 10:28 | |
| *** bogdando has joined #zuul | 10:34 | |
| bogdando | is there a way to live test by local zuul itself a change in zuul-jubs? e.g. https://review.opendev.org/#/c/696337/ | 10:35 |
| bogdando | This change looks never applied by zuul executor running my CI job | 10:35 |
| bogdando | and there is no traces for 696337 in all logs | 10:35 |
| *** tosky_ is now known as tosky | 10:38 | |
| ianw | bogdando: because configure-mirrors is included from the base job, you need to test it via base-test; see https://opendev.org/opendev/base-jobs/src/branch/master/zuul.d/jobs.yaml#L5 | 10:55 |
| bogdando | ianw: thanks! | 11:12 |
| bogdando | ianw: IIUC, I'll need to commit something to my local gerrit project copy of opendev/base-jobs for my local zuul? | 11:33 |
| bogdando | and the step 4 Propose a change to a job to reparent it to base-test is not quite clear to me. If I have my job under test is foo -> multinode -> base, what shall I reparent? | 11:35 |
| bogdando | that's multinode, right? | 11:36 |
| bogdando | like http://paste.openstack.org/show/6bLOJNKOEUDcyMvSpeOi/ ?.. | 11:39 |
| openstackgerrit | Bogdan Dobrelya (bogdando) proposed zuul/zuul-jobs master: DNM testing custom pypi mirrors (696337) https://review.opendev.org/696867 | 11:41 |
| *** jcapitao is now known as jcapitao|lunch | 12:03 | |
| bogdando | ianw: I'm sorry but do you have some example for that https://opendev.org/opendev/base-jobs/src/branch/master/zuul.d/jobs.yaml#L5 implemented as a DNM patch? | 12:05 |
| bogdando | I'm lost in understanding hose steps | 12:05 |
| bogdando | those* | 12:05 |
| *** hashar has quit IRC | 12:07 | |
| *** rfolco has joined #zuul | 12:09 | |
| *** hashar has joined #zuul | 12:10 | |
| *** jpena is now known as jpena|lunch | 12:26 | |
| Shrews | sugaar: the demo example uses the nodepool static driver (https://opendev.org/zuul/zuul/src/branch/master/doc/source/admin/examples/etc_nodepool/nodepool.yaml#L9) and uses a docker node as the target node (https://opendev.org/zuul/zuul/src/branch/master/doc/source/admin/examples/docker-compose.yaml#L89) | 12:50 |
| Shrews | sugaar: I think tristanC might be able to help you with kubernetes configuration if you're having problems with it | 12:51 |
| *** rlandy has joined #zuul | 12:57 | |
| pabelanger | ianw: bogdando: base-test will not work, unless it is also untrusted. Otherwise, you need to merge change for trusted projects | 13:12 |
| *** jcapitao|lunch is now known as jcapitao | 13:14 | |
| *** jamesmcarthur has joined #zuul | 13:15 | |
| sugaar | Shrews nice, thanks! | 13:22 |
| fungi | how we do it is to merge the alteration we want to prove to the base-test job, then optionally if the only jobs which expose the problem in question are parented through another trusted repo job to base we have to merge a duplicate of that job with a different name parented to base-test, and then finally propose a change to the affected untrusted job altering its parentage so that it chains up to base-test | 13:23 |
| sugaar | Shrews I try to workout with that info now by myself, and if I don't manage I will contact him. | 13:23 |
| *** jpena|lunch is now known as jpena | 13:24 | |
| pabelanger | fungi: issue here, is we cannot merge the role change, as it will hit production live. base / base-test playbooks don't really work if same role is in both. We need to rename role, then merge | 13:25 |
| AJaeger | bogdando: ^ | 13:26 |
| fungi | pabelanger: you need a similar role with a different name for base-test in that case | 13:27 |
| fungi | or, yeah, what you said | 13:27 |
| *** jamesmcarthur has quit IRC | 13:35 | |
| bogdando | pabelanger: I can add opendev/base-jobs as untrusted for my local zuul, just not sure of the next steps :) | 13:41 |
| bogdando | pabelanger: I think I got it. Then I will be able to use depends-on: https://review.opendev.org/696867 for example, in my local gerrit w/o merging it upstream or whatever | 13:49 |
| bogdando | so IUUC that patch only needs a copy of the base-test job ported in | 13:50 |
| bogdando | including roles and playbooks and defaults perhaps?.. | 13:50 |
| pabelanger | bogdando: is this rdoproject? | 13:51 |
| bogdando | no | 13:51 |
| pabelanger | kk | 13:51 |
| pabelanger | you should look to how we setup rdocloud, I believe configure-mirrors is located in untrusted project by default | 13:52 |
| bogdando | just doing some tripleo standalone (based on multinode zuul parent) but for fedora 31 | 13:52 |
| bogdando | thanks I'll check | 13:52 |
| bogdando | pabelanger: yes, I have zuul/zuul-jobs, which contains that configure-mirrors in untrusted | 13:53 |
| pabelanger | k, that should work now | 13:54 |
| openstackgerrit | Bogdan Dobrelya (bogdando) proposed zuul/zuul-jobs master: DNM testing custom pypi mirrors (696337) https://review.opendev.org/696867 | 13:58 |
| bogdando | mhm, that does Base jobs must be defined in config projects mean? | 14:01 |
| openstackgerrit | Bogdan Dobrelya (bogdando) proposed zuul/zuul-jobs master: DNM testing custom pypi mirrors (696337) https://review.opendev.org/696867 | 14:02 |
| tristanC | sugaar: you have to setup the ~nodepool/.kube/config similarly to the ~nodepool/.config/openstack/clouds.yaml, that is with the information and token of your provider | 14:10 |
| *** rlandy is now known as rlandy|mtg | 14:14 | |
| *** jamesmcarthur has joined #zuul | 14:17 | |
| openstackgerrit | Bogdan Dobrelya (bogdando) proposed zuul/zuul-jobs master: DNM testing custom pypi mirrors (696337) https://review.opendev.org/696867 | 14:23 |
| *** Goneri has joined #zuul | 14:41 | |
| bogdando | pabelanger, ianw, fungi: sorry, I have to give up on https://review.opendev.org/696867. I can't get how to test https://review.opendev.org/696337 manually | 14:41 |
| bogdando | If I had some real example from gerrit to compare the required steps... | 14:44 |
| *** jamesmcarthur has quit IRC | 14:49 | |
| fungi | bogdando: thanks, i missed some of the background on this so was speaking in the abstract. since there's a real-world example i'll take a look and see if i can provide a more concrete plan to test it | 14:52 |
| bogdando | fungi: thanks a lot | 14:52 |
| *** bhavikdbavishi has quit IRC | 14:53 | |
| clarkb | fwiw I think the mirror config redo plan js the proper way to address this | 14:53 |
| clarkb | then each mirrored item can be set independe tly | 14:53 |
| mordred | ++ | 15:00 |
| *** jamesmcarthur has joined #zuul | 15:03 | |
| sugaar | tristanC I am trying to do it first using minikube rather than the could provider, but do you reckon that it would be easier to go straight away to the could provider? | 15:10 |
| *** rlandy|mtg is now known as rlandy | 15:18 | |
| fungi | bogdando: while i agree with clarkb and mordred that the in-progress overhaul of mirror configuration would be a better place to spend time, the way to test that change would be to first propose a copy to zuul/zuul-jobs in roles/configure-mirrors-test and get that merged, then propose a change to opendev/base-jobs where you alter playbooks/base-test/pre.yaml to use configure-mirrors-test instead of | 15:24 |
| fungi | configure-mirrors and get that merged too, then propose a do-not-merge change to zuul/zuul-jobs which sets the parent on the unittests job to base-test and then propose a trivial do-not-merge change to some other project like zuul/zuul with a depends-on to your other do-not-merge change in zuul/zuul-jobs which gets a unit test job to run, and see if it works and exercises the pypi mirror normally but through | 15:24 |
| fungi | the new variable | 15:24 |
| fungi | and then also propose reverts of the merged changes for zuul/zuul-jobs and opendev/base-jobs | 15:25 |
| *** sgw has joined #zuul | 15:26 | |
| tristanC | sugaar: it's up to you. I guess that if you don't need cloud virtual machines, then k8s would be easier | 15:27 |
| bogdando | fungi: thanks | 15:27 |
| bogdando | > that the in-progress overhaul of mirror configuration would be a better place to spend time | 15:27 |
| bogdando | where is all that code located? | 15:27 |
| fungi | bogdando: it's sketched out in zuul-jobs documentation so far: https://zuul-ci.org/docs/zuul-jobs/mirror.html | 15:31 |
| openstackgerrit | David Moreau Simard proposed zuul/zuul master: DNM: Test zuul-stream-functional with ara 1.3 https://review.opendev.org/694622 | 15:35 |
| fungi | bogdando: in short, the configure-mirrors role was very opendev-specific and just sort of got copied over. trying to plumb more and more variables through it in an uncoordinated fashion to support uses outside opendev is going to lead to an unmaintainable mess (we might argue it already is) so the idea is to completely replace it with a new role structured for general use and flexibility | 15:36 |
| bogdando | thanks I'll try if zuul_site_mirror_info works for me | 15:38 |
| *** tflink_ is now known as tflink | 15:39 | |
| openstackgerrit | Merged zuul/nodepool master: Delete images by id https://review.opendev.org/693672 | 15:41 |
| corvus | bogdando: it's still early enough that if it doesn't, we can change the design :). if it does look good, and you wanted to make a new role (can probably copy the old role for most of it) and update it to match that design, that would be welcome. i started on something similar for the os mirrors here, but haven't finished it yet: https://review.opendev.org/677578 | 15:59 |
| *** jcapitao is now known as jcapitao|afk | 16:14 | |
| *** chkumar|ruck is now known as raukadah | 16:17 | |
| sugaar | tristanC where are stored ~nodepool/.kube/config ~nodepool/.config/openstack/clouds.yaml? are they in the container that launches nodepool? because I can't see them, or do I need to add a new user to my laptop for nodepool? | 16:22 |
| tristanC | sugaar: you need to provides those files | 16:24 |
| *** jamesmcarthur has quit IRC | 16:24 | |
| Shrews | sugaar: you might want to check how we start minikube in our zuul-jobs:install-kubernetes role: https://opendev.org/zuul/zuul-jobs/src/branch/master/roles/install-kubernetes/tasks/minikube.yaml | 16:25 |
| *** jcapitao|afk is now known as jcapitao | 16:25 | |
| *** jamesmcarthur has joined #zuul | 16:26 | |
| Shrews | i *think* minikube populates the .kube/config file for you, iirc | 16:26 |
| Shrews | you just have to tell it where it is | 16:26 |
| corvus | tristanC, clarkb: https://review.opendev.org/695050 and 2 children are ready when you have a minute (podman speculative container builds) | 16:35 |
| clarkb | k. If I make it home before dark I'll take a look. I'm about to travaerse mountain passes that have received unusual amounts of early snow | 16:38 |
| corvus | clarkb: oh, er, it's not really important enough to risk a donner party kind of situation :) | 16:40 |
| sugaar | Shrews tristanC thanks for the help, sorry it is being a bit confusing for me. I am very new with kubernetes, cloud providers and all this ecosystem and I am trying to learn about nodepool and zuul in the same time | 16:41 |
| openstackgerrit | Bogdan Dobrelya (bogdando) proposed zuul/zuul-jobs master: DNM testing custom pypi mirrors (696337) https://review.opendev.org/696918 | 16:42 |
| corvus | sugaar: :) some of us are still learning about all of those things | 16:42 |
| openstackgerrit | Bogdan Dobrelya (bogdando) proposed zuul/zuul-jobs master: DNM testing custom pypi mirrors (696337) https://review.opendev.org/696918 | 16:43 |
| Shrews | sugaar: no problem at all. but be aware that afaik, none of us are experts with minikube (i created that role that uses it with a lot of trial and error, and I'm pretty ignorant of kubernetes setup in general). but happy to help with nodepool/zuul stuff as much as possible | 16:43 |
| *** hashar has quit IRC | 16:51 | |
| openstackgerrit | Bogdan Dobrelya (bogdando) proposed zuul/zuul-jobs master: DNM testing custom pypi mirrors (696337) #2 https://review.opendev.org/696929 | 17:01 |
| bogdando | fungi: I tried https://review.opendev.org/#/c/696929/ | 17:03 |
| bogdando | and all the steps (but nothing merged yet ofc), I think it will be failing anyway as there is no base-test job defined in zuul/zuuljobs | 17:04 |
| bogdando | that patch corresponds to the step "then propose a do-not-merge change to zuul/zuul-jobs which sets the parent on the unittests job" | 17:05 |
| bogdando | I have also tried pointing out https://review.opendev.org/696918 dependency for my local gerrit, which has | 17:08 |
| bogdando | untrusted-projects: | 17:08 |
| bogdando | projects: | 17:08 |
| bogdando | - zuul/zuul-jobs | 17:08 |
| bogdando | - zuul/zuul-base-jobs | 17:08 |
| bogdando | - opendev/base-jobs | 17:08 |
| bogdando | (just to repeat the steps if it was running upstream but w/o the need to merge anything) | 17:08 |
| bogdando | and still there is no attempts to pull that 696918 in :( | 17:08 |
| bogdando | pabelanger: ^^ | 17:08 |
| pabelanger | bogdando: where do you call configure-mirrors role in your local install | 17:08 |
| pabelanger | which job? | 17:08 |
| bogdando | correction , the dependency set for my zuul was for https://review.opendev.org/#/c/696921/ | 17:09 |
| bogdando | pabelanger: there is a job on top of multinode I run for my zuul, it has depends-on: https://review.opendev.org/696921 | 17:09 |
| bogdando | that one depends on https://review.opendev.org/696918 | 17:10 |
| pabelanger | bogdando: but, we need to know where the job is defined, in your local zuul | 17:10 |
| bogdando | all that follows the steps fungi described earlier | 17:10 |
| bogdando | just that nothing has to be merged for my local zuul cuz untrusted projects ... | 17:10 |
| pabelanger | it sounds like, you call configure-mirrors from trustred playbook | 17:10 |
| pabelanger | locally | 17:10 |
| tristanC | corvus: thanks, i'll have a look shortly | 17:12 |
| bogdando | pabelanger: zuul/zuul-jobs is in untrusted-projects, and it contains configure-mirrors | 17:13 |
| *** sshnaidm is now known as sshnaidm|afk | 17:13 | |
| bogdando | opendev/base-jobs is also untrusted so I presumed I can omit the "merge" steps | 17:14 |
| pabelanger | bogdando: so, I think we are missing each other. What I am trying to understand, in your playbooks, what job calls configure-mirrors? | 17:14 |
| pabelanger | for zuul.a.c, that is our 'base' job https://github.com/ansible/ansible-zuul-jobs/blob/master/playbooks/base/pre.yaml which lives in an untrusted project | 17:15 |
| bogdando | the base I guess, my_job -> multinode -> base | 17:15 |
| pabelanger | right | 17:15 |
| pabelanger | and is base trusted? | 17:15 |
| pabelanger | I am guess yes | 17:15 |
| pabelanger | which means, you cannot do speclative testing | 17:15 |
| corvus | (it's the playbook's repo that determines whether speculative merging happens) | 17:16 |
| bogdando | pabelanger: sorry what do you mean if base is trusted? | 17:16 |
| pabelanger | bogdando: if you look at rdoproject, we creaated the idea of base-minimal (config-project) and base (untrusted) | 17:16 |
| bogdando | base is not a repo?.. | 17:16 |
| pabelanger | bogdando: base in your case, is likely a config-project: https://zuul-ci.org/docs/zuul/user/config.html#security-contexts | 17:17 |
| bogdando | pabelanger: here is untrusted et al https://pastebin.com/YDSv3G29 | 17:18 |
| pabelanger | bogdando: can you please see which job playbook as the role 'configure-mirrors' | 17:19 |
| pabelanger | Based on what you are discribing, I believe that will live in zuul-config | 17:20 |
| pabelanger | describing* | 17:20 |
| *** jcapitao has quit IRC | 17:20 | |
| bogdando | pabelanger: my local gerrit path http://localhost:9000/t/tripleo-ci-reproducer/job/base do you mean?.. https://pastebin.com/yyFWCsJn | 17:22 |
| bogdando | ok I got it, so it's trusted | 17:23 |
| pabelanger | right | 17:23 |
| pabelanger | so, none of this will work, until your change is merged | 17:23 |
| bogdando | sigh :) | 17:23 |
| pabelanger | or you move configure-mirror later in the job run, as untrusted job | 17:23 |
| pabelanger | that is what we do in zuul.a.c and pretty sure how we also setup rdoproject | 17:24 |
| bogdando | mhm, I would love to try, not sure how to do that | 17:24 |
| pabelanger | https://github.com/ansible/ansible-zuul-jobs/blob/master/zuul.d/jobs.yaml#L2 | 17:24 |
| pabelanger | is an example of untrusted 'base' that parents to trusted 'base-minimal' | 17:24 |
| bogdando | pabelanger: how gets this ansible/ansible-zuul-jobs catched up by zuul? | 17:26 |
| bogdando | via config-projects | 17:26 |
| bogdando | ? | 17:26 |
| pabelanger | bogdando: https://github.com/ansible/project-config/blob/master/zuul/tenants.yaml#L5 | 17:26 |
| pabelanger | that is our tenant config | 17:26 |
| pabelanger | ansible-zuul-jobs, is just untrusted | 17:27 |
| bogdando | pabelanger: thanks!! | 17:27 |
| pabelanger | np | 17:27 |
| *** bogdando has quit IRC | 17:31 | |
| *** Goneri has quit IRC | 17:40 | |
| *** michael-beaver has joined #zuul | 17:41 | |
| openstackgerrit | Merged zuul/zuul master: Ensure correct re-enqueue on changing gate queue https://review.opendev.org/696578 | 17:43 |
| *** jangutter has quit IRC | 17:43 | |
| *** Goneri has joined #zuul | 17:54 | |
| *** jpena is now known as jpena|off | 17:59 | |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: openshift speculative containers https://review.opendev.org/696939 | 18:01 |
| *** spsurya has quit IRC | 18:25 | |
| *** igordc has joined #zuul | 18:45 | |
| *** decimuscorvinus has quit IRC | 19:06 | |
| *** decimuscorvinus has joined #zuul | 19:07 | |
| *** sgw has quit IRC | 19:09 | |
| *** jamesmcarthur has quit IRC | 19:19 | |
| *** hashar has joined #zuul | 19:27 | |
| *** jangutter has joined #zuul | 19:44 | |
| *** jangutter has quit IRC | 19:48 | |
| *** fdegir4 has joined #zuul | 19:54 | |
| *** donnyd_ has joined #zuul | 19:54 | |
| *** persia_ has joined #zuul | 19:56 | |
| *** mugsie_ has joined #zuul | 19:56 | |
| *** pcaruana has quit IRC | 20:02 | |
| *** fdegir has quit IRC | 20:02 | |
| *** reiterative has quit IRC | 20:02 | |
| *** Guest72500 has quit IRC | 20:02 | |
| *** persia has quit IRC | 20:02 | |
| *** mugsie has quit IRC | 20:02 | |
| *** johanssone has quit IRC | 20:02 | |
| *** donnyd has quit IRC | 20:02 | |
| *** clarkb has quit IRC | 20:02 | |
| *** donnyd_ is now known as donnyd | 20:02 | |
| *** johanssone has joined #zuul | 20:03 | |
| *** clarkb has joined #zuul | 20:03 | |
| *** pcaruana has joined #zuul | 20:05 | |
| *** michael-beaver has quit IRC | 20:10 | |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: openshift speculative containers https://review.opendev.org/696939 | 20:24 |
| *** hashar has quit IRC | 20:33 | |
| ianw | corvus: i really need some help getting speculative testing working for these nodepool container jobs; do you have a minute for me to bring you up to speed? | 20:34 |
| corvus | ianw: yep | 20:35 |
| ianw | corvus: thanks; the problem for me at the moment is testing https://review.opendev.org/#/c/696486/ | 20:35 |
| ianw | it just fails with "This change depends on a change with an invalid configuration." | 20:36 |
| corvus | ianw: i'm not sure i'm convinced that we want a job running in the nodepool repo to depend on a job defined in system-config | 20:36 |
| ianw | well, where do you think the job to build the container image that will run our production nodepool containers should be? | 20:38 |
| ianw | (this may be more an infra discussion, if we're not discussing the configuration issues) | 20:39 |
| corvus | i think opendev should use the upstream zuul and nodepool container images when it switches to running in containers (i understand that's the thing that needs to change since they don't include dib). but i don't think they should depend on opendev operations. | 20:40 |
| corvus | we want zuul to be a standalone project -- it's not just something that opendev runs | 20:41 |
| corvus | we have contributors who aren't interested in contributing to opendev, and we'd like to retain them and attract more of them :) | 20:41 |
| corvus | moreover, if opendev does make its own nodepool images (as it does for gerrit) because of local ops issues, it should have the freedom to do that without impacting all zuul users | 20:42 |
| ianw | i agree, which is why i felt building the images from system-config means that they are custom images for opendev, and can make choices that may not be appropriate/useful for the generic images (such as shipping our PPA with vhd support, etc) | 20:44 |
| ianw | however, i would also like those images tested | 20:44 |
| corvus | can they be tested in system-config? | 20:45 |
| pabelanger | re: vhd support, I think we should either pull that into zuul or finally merge that code into debian directly. As, for rackspace support for nodepool, we need to install opendev repos | 20:46 |
| fungi | pabelanger: the "vhd support" being patched xen-specific image generation? if so, why not in diskimage-builder? | 20:47 |
| ianw | pabelanger: well, i basically offered to work with RAX on getting support into qemu-img for their systems (https://opendev.org/openstack/diskimage-builder/src/branch/master/doc/source/developer/vhd_creation.rst) and in so many words they were not interested | 20:47 |
| fungi | is that "support for their sustems" or "support for xen"? | 20:48 |
| fungi | s/sustems/systems/ | 20:48 |
| pabelanger | I cannot remember all the details, will read ianw link. But qemu-utils vhd support in linux doesn't work out of box | 20:48 |
| pabelanger | so, in our case we are looking to use rackspace in zuul.a.c, but need to pull in opendev ppa | 20:49 |
| ianw | pabelanger: yeah, that link has the details :) i spent quite a while, and broke the gate investigating it all :) | 20:49 |
| *** mattw4 has joined #zuul | 21:13 | |
| corvus | pabelanger: i'd be curious to hear the reasons you're looking into a packer.io nodepool builder, if you care to share | 21:20 |
| pabelanger | mostly, there is a few network appliances that have packer.io as builder. So it was an easy win to use them. However, really the workflow we need to support is boot existing qcow2 in qemu, run ansible-playbook to provision things, then compress and save. A quick look around the current builders, seems to show packer.io as leader in this (native ansible hooks). | 21:22 |
| pabelanger | we do that manually today, then manually upload into cloud. Which, isn't scaling for us | 21:23 |
| corvus | pabelanger: ah, makes sense, thanks | 21:23 |
| pabelanger | I've thought of doing it as a zuul-job, but then need to figure out upload part into nodepool-builder server, then trigger upload to cloud some how | 21:25 |
| pabelanger | with new flag to point to disk-image-create command, I'm going to POC with bash script | 21:25 |
| pabelanger | then see how to move more into nodepool correctly | 21:25 |
| corvus | pabelanger: if you want to explore the zuul job route, you can upload it to the cloud and configure nodepool-launcher to use a 'cloud-image' | 21:30 |
| pabelanger | corvus: yup, so far we haven't done that route to avoid adding nodepool credentials to zuul job but agree, on the short list as solutions. | 21:31 |
| tristanC | using a zuul-job to build and upload cloud-image would be great, but how could nodepool detect and use the new image then? | 21:40 |
| Shrews | iirc, if you're using cloud images, you just reference the image by name | 21:42 |
| pabelanger | you'd have to first delete the image, then upload new | 21:42 |
| corvus | do names have to be unique? | 21:47 |
| tristanC | perhaps nodepool could list image and pick the most recent one when if the name matches it's config? | 21:48 |
| corvus | if so, maybe we could add regex support so you can append a timestamp to the image name and then use the "latest" cloud image | 21:48 |
| corvus | tristanC: yeah, and that if names don't have to be unique | 21:48 |
| Shrews | corvus: you can also specify image id | 21:49 |
| pabelanger | no, names don't need to be unique IIRC, but nodepool will get confused if n+1 images with same name | 21:50 |
| pabelanger | well, openstacksdk | 21:50 |
| tristanC | well, not sure how the user would get the first image, but that process of using zuul-job to update nodepool image sounds like a better user experience, e.g. build logs available in zuul builds page, image update frequency with periodic pipeline, ... | 21:50 |
| corvus | if the image build can run on a *real* cloud image, that solves the bootstrap problem | 21:51 |
| pabelanger | yes, that is hold back for us too (getting image into public) | 21:51 |
| corvus | if the cloud supports renaming images, you could have a test/promote system | 21:52 |
| corvus | (even if it doesn't, you could still do that, it would just be slower) | 21:52 |
| corvus | (and have an outage window) | 21:52 |
| pabelanger | I do like the idea of test/promote of images :) | 21:55 |
| pabelanger | time to run, #dadops | 21:56 |
| *** mugsie_ is now known as mugsie | 21:59 | |
| *** pcaruana has quit IRC | 22:11 | |
| *** sgw has joined #zuul | 22:30 | |
| *** sgw has quit IRC | 22:35 | |
| *** tosky has quit IRC | 23:41 | |
| openstackgerrit | Ian Wienand proposed zuul/zuul-jobs master: [wip] copy listed siblings to docker build https://review.opendev.org/696987 | 23:53 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!