| openstackgerrit | Matthew Thode proposed openstack-infra/nodepool master: move 01-nodepool-setup to a later phase https://review.openstack.org/610188 | 00:13 |
|---|---|---|
| dmsimard | I might have a stupid idea but I wanted to share it in case it wasn't | 01:29 |
| dmsimard | Could parts of Zuul itself be delegated to playbooks ? | 01:30 |
| dmsimard | hang on, let me write an example | 01:30 |
| dmsimard | https://gist.github.com/dmsimard/baa9863e20c84538956715f4fa037963 | 01:47 |
| dmsimard | Nodepool could be a good candidate too.. there's already ansible modules for openstack, ec2, openshift, google cloud, etc. | 01:51 |
| dmsimard | For example working on the ec2 driver is kinda hard tbh, the API isn't good. | 01:54 |
| dmsimard | But there's already an ansible module that takes care of standing up and down a virtual machine in ec2 https://docs.ansible.com/ansible/2.6/modules/ec2_instance_module.html | 01:54 |
| dmsimard | same for rds, and a bunch of other amazon resources that I would never for the life of me want to write | 01:55 |
| tristanC | dmsimard: iirc, wouldn't such actions implementation be less efficient than using a library like paramiko or openstacksdk which can re-use a connection pool? | 01:55 |
| dmsimard | tristanC: we can scale the playbook execution with zuul executors, is performance really an issue ? | 01:55 |
| dmsimard | I mean, VMs are already going to take a random amount of seconds to boot and you have the ability to pre-provision them | 01:56 |
| dmsimard | you no longer need to scale launchers because they're just scheduled to run by nodepool (or zuul) on an executor | 01:57 |
| dmsimard | oh, that starts resembling a bit a discussion about a "boot" phase ahead of the "pre" phase :p | 01:57 |
| clarkb | for me the big reason for writing nodepool drivers like they are done today is expreasive control flow and control over logging and error handling | 01:58 |
| clarkb | all of that is really clunky in ansible | 01:58 |
| clarkb | so people end up writing python modules for ansible anyway | 01:59 |
| dmsimard | You can keep the amount of business logic you want | 02:00 |
| tristanC | dmsimard: in theory you can rewrite any imperative program into ansible tasks... | 02:00 |
| dmsimard | I'm talking about the idempotent creation and deletion of resources | 02:00 |
| *** dkehn has quit IRC | 02:01 | |
| dmsimard | tristanC: yeah, maybe it's a case of "if you have a hammer, everything looks like a nail" lol | 02:01 |
| clarkb | its like when people want to go from bash to "real" programming langage | 02:01 |
| clarkb | bash/ansible are great for a lot of things | 02:01 |
| dmsimard | Haha, I did a lot of perl | 02:02 |
| clarkb | eventually you need something more expressive and easier to work with at scale | 02:02 |
| dmsimard | And then I discovered python | 02:02 |
| pabelanger | oh look, peoples. I actually have a zuul question. | 02:02 |
| dmsimard | Uh oh | 02:03 |
| pabelanger | when we push a git repo to the node, how can I diff the commit between what node git repo has, and origin/master (assuming branch is master). It gerrit, it is easy, we only do a commit per patchset, but github can have multiple commits | 02:03 |
| pabelanger | git cherry might be a thing, but I'd need to add back origin/master for the repo on the node | 02:04 |
| pabelanger | and I don't think we have any reference on the repo saying, this is the current sha of master when this job ran | 02:05 |
| pabelanger | if that all makes sense | 02:05 |
| dmsimard | I understand and I have no idea | 02:06 |
| dmsimard | That is indeed problematic | 02:06 |
| clarkb | pabelanger: zuul does leave an origin/$branch ref iirc | 02:06 |
| clarkb | it might not be origin as the prefix but the pointer to diff against is there | 02:07 |
| pabelanger | let me check again | 02:08 |
| pabelanger | I only looked at git remote -v, which is origin file:///dev/null (push) | 02:09 |
| clarkb | thats orthogonal, that says where the remote is located, not what the remote refs themselves point to | 02:12 |
| clarkb | I think the worlspace setup role logs what all refs get updated | 02:12 |
| pabelanger | clarkb: okay, thanks you look to be right | 02:16 |
| pabelanger | git log origin/master..HEAD works as expected | 02:16 |
| pabelanger | should have tested first | 02:16 |
| pabelanger | before asking | 02:16 |
| *** fbo has quit IRC | 02:35 | |
| *** dkehn has joined #zuul | 02:59 | |
| *** mhu has quit IRC | 04:29 | |
| *** mhu has joined #zuul | 04:31 | |
| *** jimi|ansible has quit IRC | 05:40 | |
| openstackgerrit | Merged openstack-infra/zuul master: web: add build page https://review.openstack.org/597024 | 07:05 |
| *** ssbarnea has quit IRC | 07:26 | |
| openstackgerrit | Merged openstack-infra/zuul master: web: add job page https://review.openstack.org/597048 | 07:29 |
| *** dmellado_ has joined #zuul | 07:30 | |
| *** dmellado has quit IRC | 07:31 | |
| *** dmellado_ is now known as dmellado | 07:31 | |
| *** slaweq has joined #zuul | 07:38 | |
| *** slaweq has quit IRC | 07:48 | |
| openstackgerrit | Andreas Jaeger proposed openstack-infra/zuul-jobs master: Fix RST formatting https://review.openstack.org/610198 | 08:36 |
| tristanC | if you ever wondered what would look like the zuul logo in 3d, here is a demo: https://www.shadertoy.com/view/XldfW8 :-) | 09:11 |
| *** dkehn has quit IRC | 09:17 | |
| *** evrardjp_ has joined #zuul | 10:19 | |
| *** swest has quit IRC | 10:20 | |
| *** evrardjp has quit IRC | 10:22 | |
| *** swest has joined #zuul | 11:15 | |
| mordred | tristanC: nice! | 12:26 |
| *** dkehn has joined #zuul | 12:30 | |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul master: Use zuul/web/static in setup_hook https://review.openstack.org/609764 | 12:40 |
| mordred | tristanC: on the job page - did you think yet about making the context field have a link to the browseable git repo for the file? I'm guessing it would involve being able to configure a browse-this-file link pattern for each source | 12:49 |
| mordred | tristanC, corvus: interesting ... check out http://zuul.openstack.org/job/devstack-tox-functional | 12:54 |
| mordred | tristanC: stable/pike is in the branch list twice - and I can't see any difference | 12:55 |
| *** bhavikdbavishi has joined #zuul | 13:05 | |
| tristanC | mordred: it also appears twice in http://zuul.openstack.org/api/job/devstack-tox-functional | 13:28 |
| tristanC | mordred: also, have you seen https://review.openstack.org/#/c/604264/ ? it shows an issue regarding to how project-template configs are currently returned by the api | 13:29 |
| tristanC | mordred: for links, we may needs an api/connection to return base_url/giturl, and adds the connection name to the source_context | 13:30 |
| *** bhavikdbavishi has quit IRC | 13:31 | |
| corvus | that's weird. we may need the repl to figure out where that came from. or maybe run an instance of devstack-tox-functional and see if the inheritance_path gives any clues. | 14:28 |
| dmsimard | tristanC: now I need a 3d printed version :p | 14:28 |
| corvus | mordred: \o/ the zuul-web image works! | 14:30 |
| corvus | next week i'll add that to the quickstart, and then we'll be gating on it | 14:30 |
| mordred | corvus: woot! | 14:44 |
| *** smyers has quit IRC | 14:58 | |
| *** dkehn has quit IRC | 15:17 | |
| *** dkehn has joined #zuul | 15:20 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/nodepool master: Update pypi metadata https://review.openstack.org/557738 | 16:07 |
| *** evrardjp_ has quit IRC | 19:07 | |
| *** evrardjp_ has joined #zuul | 19:09 | |
| openstackgerrit | Andreas Jaeger proposed openstack-infra/zuul-jobs master: Use :default: for emit-ara-html/README.rst https://review.openstack.org/610286 | 20:11 |
| *** evrardjp_ has quit IRC | 20:11 | |
| *** evrardjp_ has joined #zuul | 20:12 | |
| *** jesusaur has quit IRC | 23:24 | |
| *** jesusaur has joined #zuul | 23:29 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!