| *** dmsimard is now known as dmsimard|off | 02:17 | |
| *** bhavik1 has joined #zuul | 03:32 | |
| *** bhavik1 has quit IRC | 03:33 | |
| *** qwc has quit IRC | 11:30 | |
| *** qwc has joined #zuul | 11:31 | |
| SpamapS | The more I play with zuul, the more I want a zuul-ansible-playbook entry point that will setup the roles path for me and execute a playbook in bwrap sort of the same way the executor would. Would be very useful for local playbook iterating. | 17:00 |
|---|---|---|
| dmsimard|off | SpamapS: how does it do that without a running scheduler that loads the config from 2000 repos ? :) | 17:23 |
| dmsimard|off | FWIW your idea runs along the same lines of what I've been thinking about doing to test trusted playbooks/jobs | 17:25 |
| SpamapS | dmsimard|off: I believe the config needs to be accessible via REST API | 17:27 |
| SpamapS | so just like we ask the scheduler for keys, we can ask it for configs. | 17:27 |
| dmsimard|off | We can't speculatively test many things and it's bitten more than once.. It's hard to come up with ideas that don't involve somehow running a nested zuul that runs the patch(es) you're testing | 17:27 |
| dmsimard|off | but the config is massive, isn't it ? | 17:27 |
| SpamapS | so 'zuul-ansible-playbook -c https://zuulv3.openstack.org/openstack playbooks/foo.yaml' | 17:28 |
| dmsimard|off | I mean it's probably not going to be the size of openstack for other deployments, sure, but how do you expose that without killing the server ? | 17:28 |
| SpamapS | dmsimard|off: If-None-Match and a good HTTP cache? ;-) | 17:28 |
| dmsimard|off | hm, I'd rather write a "compiled" version of the config to disk each time the config is reloaded and then have the API call return that file so that you don't poke the actual server instead | 17:29 |
| SpamapS | That's basically the same thing done on the backend instead of the frontend. :) | 17:30 |
| SpamapS | lazy-load vs. pre-cache | 17:30 |
| dmsimard|off | http cache and headers are not part of zuul though | 17:30 |
| SpamapS | Yeah, it could be done pretty easily in zuul. I like it. | 17:31 |
| SpamapS | Assuming it can be serialized relatively quickly. | 17:31 |
| SpamapS | Anyway, so many good features, so little time. :) | 17:31 |
| * dmsimard|off nods | 17:31 | |
| SpamapS | For now, I have a script that just sets ANSIBLE_ROLES_PATH and a static zuul.json that I -e @ in | 17:31 |
| dmsimard|off | I was supposed to have a beta of ara 1.0 in september | 17:31 |
| dmsimard|off | so much for that, I'm maybe half done | 17:31 |
| dmsimard|off | SpamapS: the other thing about your idea is running on localhost (executor) vs remote | 17:32 |
| SpamapS | dmsimard|off: yeah, no groups :) | 17:33 |
| SpamapS | right now I just -i localhost, | 17:33 |
| SpamapS | and if I need an inventory, I invent one. | 17:33 |
| dmsimard|off | ಠ_ಠ | 17:34 |
| SpamapS | honestly most of the time I just want to make sure the syntax checks out so I don't embarass myself when pushing up PR's/patches. :) | 18:18 |
| SpamapS | jeblair: so.. interesting problem I just ran into. | 18:51 |
| SpamapS | jeblair: we have a local mirror of all the openstack projects. I'd like to push those into some of our local jobs. But the zuul.yaml in those repos gets loaded, and fails, because I don't have all the other jobs/repos. Thoughts? | 18:52 |
| SpamapS | We may need to have some way in the tenant config to prevent loading in-repo configs for a repo. | 18:57 |
| *** jkilpatr has quit IRC | 19:00 | |
| *** jkilpatr has joined #zuul | 19:06 | |
| *** jkilpatr has quit IRC | 23:35 | |
| dmsimard|off | SpamapS: I was wondering about that. For example, once our zuul v3 is up, we might want to add zuul-jobs to benefit from playbooks/roles/jobs that are defined there, but that also means loading the zuul.yaml, which means there can be syntax errors (for our zuul) | 23:38 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!