| *** jimi|ansible has quit IRC | 00:48 | |
| openstackgerrit | David Moreau Simard proposed openstack-infra/nodepool master: WIP: Implement an Amazon EC2 driver https://review.openstack.org/535558 | 01:18 |
|---|---|---|
| SpamapS | so, before I solve it in an ugly way.. has anybody solved /usr/bin/python missing in an elegant way? | 01:18 |
| dmsimard | SpamapS: on a remote node ? | 01:18 |
| SpamapS | Yes.. I want to use a stock image that doesn't have python2 installed | 01:19 |
| dmsimard | SpamapS: https://docs.ansible.com/ansible/2.6/modules/raw_module.html | 01:19 |
| SpamapS | and Ansible has really screwed this up IMO (they really should try /usr/bin/python3 first and then fall back to /usr/bin/python) | 01:19 |
| SpamapS | dmsimard: that won't work for zuul | 01:19 |
| SpamapS | zuul starts by doing something with -m setup | 01:19 |
| SpamapS | which 'splodes because of missing /usr/bin/python | 01:19 |
| dmsimard | SpamapS: oh no | 01:19 |
| dmsimard | We can fix that | 01:20 |
| SpamapS | Well, I was just going to go ahead and set ansible_python_interpreter somewhere and make sure it gets set on every invocation | 01:21 |
| SpamapS | One way to solve this is to make sure the job variables and site variables get plumbed through to that call to setup | 01:21 |
| dmsimard | SpamapS: I wonder if you could boot with user-data and make something install before opening up ssh | 01:21 |
| dmsimard | SpamapS: s/install/run/ .. like ln -s /usr/bin/python3 /usr/bin/python or something | 01:22 |
| dmsimard | SpamapS: dunno, there's probably a lot of ways to work around it | 01:22 |
| clarkb | spamaps ya ansible should do more imo | 01:23 |
| *** jiapei has joined #zuul | 01:23 | |
| SpamapS | dmsimard: yeah that's not a bad idea | 01:23 |
| SpamapS | but ... ansible.. totally blowing it IMO | 01:23 |
| SpamapS | and ... I'd submit a patch.. but... wow.. they seem busy over there | 01:24 |
| dmsimard | SpamapS: not disagreeing, trying to look through their issues if they have something | 01:24 |
| dmsimard | SpamapS: is that a stock AMI on ec2 ? | 01:27 |
| SpamapS | I mean I don't even need userdata, I can just snapshot an instance after doing that and use it. But... yeah. | 01:27 |
| SpamapS | dmsimard: yeah, ubuntu 16.04 and 18.04 both ship without /usr/bin/python | 01:27 |
| dmsimard | SpamapS: ok, let me ping around | 01:27 |
| SpamapS | I'm just making my own AMI right now | 01:27 |
| SpamapS | by snapping one that's already booted with the symlink | 01:28 |
| SpamapS | but.. seems like we should make zuul figure this out on its own. | 01:28 |
| SpamapS | And ultimately, ansible. | 01:28 |
| dmsimard | SpamapS: I'm not sure what's the big deal in making Ansible attempt to use /usr/bin/python3 (or /usr/bin/python2) if /usr/bin/python is not present... | 01:29 |
| clarkb | fwiw chances are you want to build your own images anyway? but ya ansible gives up really quivk imo | 01:29 |
| SpamapS | Even if I do build my own | 01:30 |
| SpamapS | I want to use python3 | 01:30 |
| SpamapS | always | 01:30 |
| SpamapS | and /usr/bin/python != python3 | 01:30 |
| clarkb | ya zuul uses python2 for some reason that may not be necessary anymore | 01:30 |
| SpamapS | That's already pretty much been stipulated by the python world. :-P | 01:30 |
| clarkb | I forget the reason though | 01:30 |
| SpamapS | I don't see zuul setting it at all | 01:30 |
| SpamapS | It's just the default value | 01:31 |
| SpamapS | /usr/bin/python | 01:31 |
| SpamapS | and that's fine.. I kind of feel like zuul should make it easy to set ansible_python_interpreter per job or label. | 01:31 |
| SpamapS | In fact I had hoped vars: { ansible_python_interpeter: /usr/bin/python3 } would work | 01:32 |
| SpamapS | but it doesn't, because we don't set job vars on that -m setup run | 01:32 |
| clarkb | https://git.openstack.org/cgit/openstack-infra/zuul/tree/zuul/executor/server.py#n1509 | 01:32 |
| clarkb | it sets it there at least | 01:32 |
| SpamapS | Ah that's very interesting | 01:32 |
| clarkb | git log/blame on that line will likely say why | 01:32 |
| clarkb | I know there was a reason though | 01:32 |
| SpamapS | well then I'll just install python2 and deal with it | 01:33 |
| dmsimard | SpamapS: added a comment in https://github.com/ansible/ansible/issues/19605#issuecomment-417160811 | 01:35 |
| dmsimard | worth a shot :) | 01:36 |
| *** rlandy|bbl is now known as rlandy | 01:37 | |
| *** rlandy has quit IRC | 01:42 | |
| SpamapS | dmsimard: awsome thanks | 01:50 |
| *** bhavikdbavishi has joined #zuul | 02:54 | |
| *** jiapei has quit IRC | 03:33 | |
| *** bhavikdbavishi has quit IRC | 03:34 | |
| *** snapiri has joined #zuul | 05:59 | |
| *** openstackgerrit has quit IRC | 06:07 | |
| *** pcaruana has joined #zuul | 07:14 | |
| *** goern has joined #zuul | 07:19 | |
| *** hashar has joined #zuul | 07:35 | |
| *** hwoarang has joined #zuul | 07:49 | |
| *** jpena|off is now known as jpena | 07:54 | |
| *** electrofelix has joined #zuul | 08:21 | |
| *** panda is now known as panda|rover | 08:52 | |
| *** jesusaur has quit IRC | 09:27 | |
| *** jesusaur has joined #zuul | 09:34 | |
| *** hashar is now known as hasharLunch | 10:11 | |
| *** hwoarang has quit IRC | 11:20 | |
| *** hwoarang has joined #zuul | 11:20 | |
| *** hasharLunch is now known as hashar | 11:21 | |
| *** jpena is now known as jpena|lunch | 11:27 | |
| *** sshnaidm is now known as sshnaidm|afk | 11:28 | |
| *** threestrands has quit IRC | 11:38 | |
| *** TheAceOfHearts has joined #zuul | 11:41 | |
| *** hwoarang has quit IRC | 11:46 | |
| *** hwoarang has joined #zuul | 11:46 | |
| *** gouthamr has quit IRC | 11:49 | |
| *** dmellado has quit IRC | 11:50 | |
| *** openstackgerrit has joined #zuul | 12:09 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: web: add config-errors notifications drawer https://review.openstack.org/597147 | 12:09 |
| *** jpena|lunch is now known as jpena|off | 12:31 | |
| *** jpena|off is now known as jpena | 12:32 | |
| *** rlandy has joined #zuul | 12:37 | |
| *** myoung|pto is now known as moyung | 13:03 | |
| *** moyung is now known as myoung | 13:03 | |
| openstackgerrit | Simon Westphahl proposed openstack-infra/nodepool master: Update static nodes in Zookeeper on config change https://review.openstack.org/598156 | 13:11 |
| dmsimard | tristanC, SpamapS: re: ec2 driver, I sent PS11/12 which adds support for profile selection (i.e, something other than [default] in .aws/credentials), works around assumptions around block device mapping and otherwise tries to bring the driver more in line with what the openstack driver looks like | 13:13 |
| dmsimard | the required config looks a bit different now and I've updated the fixture to reflect that https://review.openstack.org/#/c/535558/10..12/nodepool/tests/fixtures/config_validate/good.yaml | 13:14 |
| openstackgerrit | Simon Westphahl proposed openstack-infra/nodepool master: Update static nodes in Zookeeper on config change https://review.openstack.org/598156 | 13:18 |
| swest | Shrews: tristanC: ^ I made a proposal for updating static nodes in Zookeeper if the config changed. Would be great if you could give some feedback | 13:21 |
| Shrews | swest: left a comment | 13:27 |
| tristanC | Shrews: swest: left a comment too, i wonder why static config change aren't handled by the config object __eq__ checks instead? | 13:28 |
| tristanC | dmsimard: works for me. I guess we now need to add tests for getting approved in nodepool? | 13:30 |
| dmsimard | tristanC: there's still a lot of work, I'll add a list of todo in the commit message | 13:31 |
| dmsimard | tristanC: quota management is one, yes -- but also diskimage, processes like deleting leaked nodes, etc | 13:32 |
| tristanC | dmsimard: couldn't these be added as follow-up? | 13:33 |
| *** samccann has joined #zuul | 13:35 | |
| dmsimard | tristanC: I think quota management and diskimage management might be appropriate for a second iteration but things like deleting leaked nodes aren't | 13:37 |
| dmsimard | tristanC: I also suck at writing tests so I hope someone can help with that | 13:37 |
| dmsimard | SpamapS, clarkb: #ansible-devel confirmed that they would also like Ansible to attempt to use something other than /usr/bin/python if it's not there. They've already given a few shots at it but were not happy with the implementations, it's likely to be in the 2.8 roadmap | 13:41 |
| *** toabctl has joined #zuul | 13:44 | |
| swest | tristanC: not quite sure what you mean with "... handled by the config object __eq__ checks instead", since the node config is a dict. | 13:51 |
| tristanC | swest: for openstack, the provider is removed if the config eq failed, e.g. https://git.zuul-ci.org/cgit/nodepool/tree/nodepool/provider_manager.py#n55 | 13:52 |
| tristanC | and there: https://git.zuul-ci.org/cgit/nodepool/tree/nodepool/driver/openstack/config.py#n92 | 13:53 |
| tristanC | that's how labels setting update get applied during reconfiguration | 13:54 |
| swest | tristanC: I think the provider manager and __eq__ check is fine. Adding a new pool also works. The problem is that if the config changes and a new provider is created the existing static nodes are not updated. | 14:03 |
| *** jimi|ansible has joined #zuul | 14:08 | |
| *** sshnaidm|afk has quit IRC | 14:10 | |
| *** sshnaidm|afk has joined #zuul | 14:19 | |
| *** sshnaidm|afk is now known as sshnaidm | 14:20 | |
| *** sdoran has joined #zuul | 15:03 | |
| *** pcaruana has quit IRC | 15:12 | |
| openstackgerrit | Sorin Sbarnea proposed openstack-infra/zuul master: Display correct cursor when hovering over patchset header https://review.openstack.org/598222 | 16:12 |
| *** shachar has joined #zuul | 16:12 | |
| *** snapiri has quit IRC | 16:14 | |
| *** jimi|ansible has quit IRC | 16:16 | |
| *** rlandy is now known as rlandy|afk | 16:19 | |
| *** gouthamr has joined #zuul | 16:49 | |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Provide per-project ssh key to executor https://review.openstack.org/597727 | 16:51 |
| SpamapS | dmsimard: thanks for the follow-up! | 17:23 |
| SpamapS | dmsimard: I'll pull in ,12 today | 17:24 |
| *** jpena is now known as jpena|away | 17:27 | |
| dmsimard | SpamapS: neat, let me know if you run into anything -- it's nowhere finished yet | 17:30 |
| SpamapS | dmsimard: would be nice to get something merged soon | 17:41 |
| dmsimard | agreed but I don't think it matches the quality of something that can be released | 17:41 |
| dmsimard | (yet) | 17:41 |
| *** dmellado has joined #zuul | 17:44 | |
| SpamapS | dmsimard: It's beta-or-so | 17:45 |
| SpamapS | ;) | 17:45 |
| SpamapS | I'd really also like to figure out how we can leverage nodepool-builder in AWS | 17:45 |
| SpamapS | I was looking at the Denver PTG schedule | 17:47 |
| SpamapS | and I didn't see zuul explicitly mentioned | 17:47 |
| dmsimard | believe fungi or clarkb were looking at putting a pad together | 17:48 |
| clarkb | SpamapS: zuul decided to not have an official on calendar presence at the PTG aiui | 17:48 |
| SpamapS | mmk | 17:48 |
| clarkb | but I'm sure there will be carvable corners for such things | 17:49 |
| clarkb | as for AWS support by nodepool builder I Think dib can already build amis (but may need better docs around that), then its just a matter of having the builder speak the upload api? | 17:49 |
| *** electrofelix has quit IRC | 17:50 | |
| dmsimard | SpamapS: I don't have time to spend on nodepool-builder/diskimage support in the short term unfortunately (really tight time constraints) | 17:50 |
| SpamapS | dmsimard: yeah I was looking into it yesterday | 17:50 |
| SpamapS | then gave up and just snapshotted an instance and used that ami | 17:50 |
| dmsimard | haha | 17:50 |
| dmsimard | yeah it's a non-negligible amount of work that I don't have the bandwidth -- our image builds are just done out of band right now (or we pick available AMIs).. it works for the time being so I'll circle around back to it later if I can | 17:51 |
| dmsimard | I don't even know what's special about AMIs vs regular qcow2/raw images we're building today | 17:52 |
| clarkb | dmsimard: the disk and kenrel are separate iirc, otherwise not much | 17:52 |
| *** hashar is now known as hasharDinner | 18:00 | |
| dmsimard | I am tagging ARA 0.16.0 for release, no problems spotted in automated and manual testing | 18:09 |
| SpamapS | Yeah there's a few scripts out there that will slurp the boot bits out of a qcow2 and make an ami | 18:35 |
| *** toabctl has quit IRC | 18:51 | |
| *** rlandy|afk is now known as rlandy | 19:05 | |
| *** toabctl has joined #zuul | 19:07 | |
| SpamapS | silly question: how do I tie a template to every single project? | 19:11 |
| dmsimard | SpamapS: there was a regex feature that Tobias was working on, not sure if it landed | 19:15 |
| dmsimard | SpamapS: looks like it has https://review.openstack.org/#/c/535713/ | 19:15 |
| SpamapS | dmsimard: oh that's perfect | 19:17 |
| clarkb | SpamapS: https://git.openstack.org/cgit/openstack-infra/project-config/tree/zuul.d/projects.yaml#n4 an example of it in use | 19:20 |
| SpamapS | clarkb: yeah that's exactly what I need | 19:21 |
| SpamapS | I have a silly github-only pipeline that turns approved reviews into labels so that people can use labels for self-approve... but up until now I was just having to hang it off each project. | 19:21 |
| *** jpena|away is now known as jpena|off | 19:38 | |
| *** hasharDinner is now known as hashar | 19:56 | |
| *** mugsie has quit IRC | 20:25 | |
| *** TheAceOfHearts has quit IRC | 20:35 | |
| *** mugsie has joined #zuul | 20:39 | |
| *** harlowja has joined #zuul | 20:43 | |
| *** samccann has quit IRC | 21:11 | |
| openstackgerrit | Sorin Sbarnea proposed openstack-infra/zuul master: Display correct cursor when hovering over patchset header https://review.openstack.org/598222 | 21:13 |
| ssbarnea | if anyone can approve that one liner ^^ css fix it would be great. I tested using styler and works well. | 21:16 |
| ssbarnea | rlandy:if you can review https://review.openstack.org/#/c/571176/ it would be great, is one line, and zero risk. | 21:20 |
| rlandy | voted | 21:22 |
| ssbarnea | thanks | 21:22 |
| *** dmellado has quit IRC | 21:31 | |
| *** dmellado has joined #zuul | 21:31 | |
| dkehn | We have and instance of zuul running. The zuul-web client is serving on port 9000. We can see the tenant page in a browser. When we point a github webhook to http://<hostname>:9000/api/connection/github we don anything on the server side logs for zuul nor does the payload deliver correctly from github, any advise?'t see | 21:38 |
| pabelanger | https://zuul-ci.org/docs/zuul/admin/github_setup.html says it is http://<IP ADDRESS>:9000/api/connection/github/payload | 21:40 |
| pabelanger | assuming github is the connection name | 21:40 |
| dkehn | sorry I forgot the word payload | 21:42 |
| dkehn | and we have DNS, I'm assuming that will work or must it be IP_ADDRESS | 21:43 |
| pabelanger | dkehn: and zuul-web is running? | 21:44 |
| dkehn | yes | 21:44 |
| *** dkehn is now known as dkehn_ | 21:46 | |
| *** myoung has quit IRC | 21:47 | |
| *** dkehn has joined #zuul | 21:49 | |
| pabelanger | dkehn_: what connection name did you use for github.com | 21:50 |
| dkehn | pabelanger: yes | 21:51 |
| dkehn | pabelanger: sorry had to switch to different system | 21:51 |
| pabelanger | np, I was asking what was the name use setup in zuul.conf for github.com connection | 21:52 |
| pabelanger | [connection github] | 21:52 |
| dkehn | [connection github] | 21:54 |
| dkehn | sshkey = /var/lib/zuul/.ssh/id_rsa | 21:54 |
| dkehn | driver = github | 21:54 |
| dkehn | api_token = <secret> | 21:54 |
| dkehn | webhook_token = <secret> | 21:54 |
| dkehn | git_host = github.com | 21:54 |
| pabelanger | and 9000 is open in on firewall? | 21:55 |
| dkehn | yes its open | 21:56 |
| pabelanger | you should be able to use curl --request POST with some dummy json to the URL and see the request in zuul debug logs | 21:56 |
| dkehn | iptables confirmed | 21:56 |
| dkehn | trying | 21:57 |
| *** TheAceOfHearts has joined #zuul | 21:59 | |
| dkehn | we seem to be getting a 400 Bad Request | 21:59 |
| pabelanger | for example | 21:59 |
| pabelanger | http://zuul.openstack.org/api/connection/github/payload | 22:00 |
| pabelanger | is what we have in openstack-infra I believe | 22:00 |
| pabelanger | but, because it doesn't respond to GET, get 500 | 22:00 |
| pabelanger | dkehn: check debug logs in zuul-web | 22:00 |
| pabelanger | I have to run now for some food, will catch up on backscroll | 22:00 |
| pabelanger | good luck | 22:01 |
| *** TheAceOfHearts has quit IRC | 22:01 | |
| dkehn | we can't seem to get zuul-web logging | 22:01 |
| *** TheAceOf_ has joined #zuul | 22:01 | |
| *** TheAceOf_ is now known as TheAceOfHearts | 22:04 | |
| corvus | dkehn: another useful resource is in github, you can see the response github receives to its payload requests | 22:07 |
| dkehn | corvus: thanks | 22:12 |
| *** hashar has quit IRC | 22:16 | |
| *** sshnaidm is now known as sshnaidm|off | 22:35 | |
| SpamapS | hm | 22:35 |
| SpamapS | I think something changed in the github driver subtly (probably in release notes I know). I have a pipeline that tags stuff on approved reviews, but it won't even admit a change into its queue until the change satisfies Github's "can this merge" criteria (it's missing required checks). I'd like to disable that.. anybody know how? | 22:36 |
| corvus | SpamapS: i take it that pipeline doesn't run the checks (ie, this isn't gate, this is something else?) | 22:39 |
| SpamapS | corvus: http://paste.openstack.org/show/729189/ | 22:40 |
| SpamapS | that's the entirety of the pipeline | 22:40 |
| SpamapS | And then I have a template that attaches a noop in that pipeline to every project | 22:40 |
| SpamapS | (It has an unlabel-on-push counterpart too) | 22:41 |
| corvus | SpamapS: is it 'dependent' for a reason? | 22:41 |
| SpamapS | corvus: no reason I can think of | 22:41 |
| SpamapS | I probably just thought "seems legit" | 22:41 |
| corvus | SpamapS: i think independent doesn't perform the mergability check, so if you switch that you should be good | 22:41 |
| SpamapS | yeah that makes sense | 22:41 |
| openstackgerrit | Merged openstack-infra/zuul master: Add missing parentheses to title invocation https://review.openstack.org/595361 | 22:48 |
| *** TheAceOfHearts is now known as TheAceOfHearts_ | 23:22 | |
| openstackgerrit | Ian Wienand proposed openstack-infra/zuul-jobs master: Add guidelines on variable naming https://review.openstack.org/595920 | 23:28 |
| dmsimard | tristanC: you know what ? | 23:32 |
| tristanC | SpamapS: if adding ec2 to nodepool tree is possible, how are we going to manage its requirements? shouldn't we use "boto3 [aws]" extra flag? | 23:32 |
| clarkb | tristanC: is there a reason to not just alwaysinstall boto if nodepool supports ec2? | 23:34 |
| dmsimard | tristanC: writing all this python code with boto and the nodepool business logic kind of sucks tbh. I wonder if we could just use Ansible as the "driver" ? Ansible has modules to create VMs in AWS idempotently and manage them. It's complicated and they got it all figured out. | 23:35 |
| tristanC | clarkb: to reduce complexicity. aws usere shouldn't have to pull openstack sdk and vice-versa | 23:35 |
| clarkb | tristanC: I mean I pip install and don't think about it typically | 23:35 |
| clarkb | (even less thinking if using the pbrx images) | 23:36 |
| dmsimard | tristanC: it's kind of the same question that comes up in packaging... Like for example for Glance/Cinder/Neutron, we don't install all the backend packages | 23:37 |
| dmsimard | Distributions don't typically install everything but they might have a sane default, though | 23:37 |
| tristanC | clarkb: works for me too :) | 23:38 |
| clarkb | dmsimard: there has been talk of potentially supporting a generic ansible provisioner then if ansible supports it so would nodepool | 23:38 |
| tristanC | dmsimard: actually, the nodepool package in fedora has a sub-package for openstack that requires openstacksdk | 23:38 |
| clarkb | dmsimard: the trouble comes when you need to cleanup leaked resources as with openstack (unsure if you run into similar problems with aws) and need more business logic in nodepool itself | 23:39 |
| dmsimard | clarkb: exactly, that's the point, then we don't need to write all the provisioning logic | 23:39 |
| clarkb | I know gce suffers the same issues | 23:39 |
| dmsimard | Yeah.. it's not perfect but it's fixable | 23:39 |
| tristanC | dmsimard: the issue with using ansible is speed, using a sdk seems more efficient for nodepool | 23:39 |
| dmsimard | tristanC: I'm not sure the tradeoff in performance vs maintability/feature is worth the hassle | 23:40 |
| clarkb | I'm personally a fan of the proper sdk implementations myself just based on the amount of work it takes to make openstack clouds happy and hearing similar problems from users of other clouds | 23:40 |
| clarkb | basically ansible will cut it for when you boot 2 VMs a day, not 25k | 23:40 |
| dmsimard | Really? Why? | 23:40 |
| clarkb | because you leak all the things, ports, volumes, floating IPs | 23:41 |
| clarkb | and you need to have software that manages that | 23:41 |
| SpamapS | It's a neat idea to make a generic ansible driver..... | 23:41 |
| dmsimard | That's the purpose of nodepool, no ? | 23:41 |
| dmsimard | I'm not saying let's get rid of nodepool | 23:41 |
| SpamapS | but as we go more toward "thing pool" and less "node pool", I think native drivers are going to make more sense. | 23:42 |
| dmsimard | I'm saying nodepool should use ansible to provision things instead of writing code for that | 23:42 |
| clarkb | dmsimard: yup, I just don't know how you'd effectively do that with ansible in the middle? ansible list all ports, ok now show all of them and figure out which ones are active and attached to valid instance, Now go back through and delete the rest. Expressing that in python is far easier | 23:42 |
| dmsimard | The aws modules are like 1k lines long each and there's dozens DOZENS | 23:42 |
| SpamapS | dmsimard: yeah I don't agree and I think that's going to end up being just as hard as writing good drivers. | 23:42 |
| clarkb | basically its easier to express the lifecycle and associated exceptions in a proper programming language | 23:42 |
| SpamapS | ^^ that | 23:43 |
| dmsimard | clarkb: the leaking cleanup doesn't have to be ansible | 23:43 |
| SpamapS | and aws's resource based API is pretty close to shade. | 23:43 |
| tristanC | sdk also improve connection pool and keep a single authenticated session | 23:43 |
| SpamapS | (the low level one is just as awful as low level openstack) | 23:43 |
| dmsimard | The tradeoff would be to be able to support all the things Ansible supports | 23:43 |
| dmsimard | Pros and cons :) | 23:44 |
| clarkb | the associated tradeoff is the support is only as good as ansibles which ime is honestly not always that great (for aws its prbably fine) | 23:44 |
| SpamapS | I think the struggle right now is unwinding the assumptions we've been able to build up with just two very different drivers (static and openstack) | 23:44 |
| SpamapS | dmsimard: but that' a falacy, because we will still have to write ansible playbooks that work the way your playbook driver expects, and some ansible modules are nothing like others. | 23:45 |
| clarkb | in particular backward incompatible changes show up every release and a lot of the provisioning modules do not have stability guaruntees like the "core" of ansible last I checked | 23:45 |
| SpamapS | In fact frankly, I don't think Ansible is very good as a provisioner. | 23:45 |
| SpamapS | I'd choose terraform before I'd choose Ansible, for that. | 23:45 |
| SpamapS | but then I look at that and I think.. for this narrow use case.. 1000 lines of python is probably better than 1000 lines of ansible. | 23:46 |
| dmsimard | SpamapS: sure | 23:46 |
| clarkb | (looks like ec2 is makred stableinterface at least) | 23:46 |
| dmsimard | Right tool for the right job | 23:46 |
| dmsimard | If terraform is better, let's use it :D | 23:46 |
| dmsimard | (I've been meaning to try it) | 23:46 |
| dmsimard | I know there's an Ansible module for terraform but that's about it | 23:47 |
| clarkb | (azure is preview) | 23:47 |
| SpamapS | It's called command: ;-) | 23:48 |
| *** TheAceOfHearts_ has quit IRC | 23:48 | |
| * SpamapS isn't a fan of ansible modules that just wrap CLI tools that have perfectly reasonable CLI's | 23:49 | |
| SpamapS | If your command sucks at idempotency, ok, wrap it. | 23:49 |
| clarkb | as a side note, the openstack modules are basically nodepool's grandchildren (nodepool -> shade -> os_* mdules in ansible). At least in that case we had to do the work in nodepool first and there isn't much incentive to add more layers to it in nodepool | 23:51 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!