| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2018-04-05 | |||
| 17:11:06 | mordred | it should be noted also that script is only doing things with interface='public' | |
| 17:11:22 | mordred | dtantsur: yah | |
| 17:11:37 | dtantsur | I already have a few hacks in https://github.com/dtantsur/rust-openstack/blob/master/src/common/protocol.rs#L88-L131 | |
| 17:11:43 | dtantsur | and I will certainly have more... | |
| 17:11:57 | mordred | dtantsur: yah | |
| 17:12:20 | mordred | dtantsur: in fact, I've been thinking that it would be good to document the various hacks that we do in os-client-config/keystoneauth | |
| 17:12:29 | dtantsur | I'd love it | |
| 17:12:39 | mordred | since it is *completely* impossible to consume the apis generally without applying several client-side hacks | |
| 17:12:49 | dtantsur | as to this code, I guess I'll have to update it to support several major versions.... brrrrr | |
| 17:13:12 | mordred | dtantsur: for instance, glance frequenty returns invalid urls in its version discovery documents because it's easy for admins to misconfigure it ... | |
| 17:13:42 | mordred | dtantsur: so it'll return the internal address - we fix that by taking what we find in the version discovery doc and combining it with what's in the catalog | |
| 17:13:54 | mordred | dtantsur: seems like the sort of thing we shoudl document for other people :) | |
| 17:13:56 | dtantsur | O.U.C.H | |
| 17:14:07 | mordred | (we do that for *all* version discovery documents, not just glance) | |
| 17:14:21 | dtantsur | kind of the same as the https hack you see in the link, right :) | |
| 17:14:27 | mordred | but also- similar hacks have to be employed for pagination links | |
| 17:14:30 | mordred | dtantsur: yup | |
| 17:14:39 | dtantsur | pagination has been a big pain for me so far | |
| 17:14:59 | dtantsur | I'm not even using pagination links, which is probably bad | |
| 17:15:01 | mordred | dtantsur: we've got an impl in sdk that handles all of the cases you might want to look at | |
| 17:15:11 | dtantsur | link? :) | |
| 17:15:11 | mordred | dtantsur: yah - using hte pagination links actually make it much nicer | |
| 17:15:14 | mordred | coming ... | |
| 17:15:38 | dtantsur | mordred: I'll have to rework some internals of rust-openstack to use them | |
| 17:16:20 | mordred | dtantsur: starting here: http://git.openstack.org/cgit/openstack/openstacksdk/tree/openstack/resource.py#n887 | |
| 17:16:39 | dtantsur | thanks! | |
| 17:16:54 | dtantsur | my other problem is that I only have one production cloud to test on (namely, RDO cloud) | |
| 17:17:01 | mordred | dtantsur: there are a few different forms of pagination links across the services - to my knowledge, that code appropriately consumes all of them without resorting to making extra calls needlessly | |
| 17:17:09 | mordred | dtantsur: that is a problem we can solve :) | |
| 17:17:42 | dtantsur | rly? | |
| 17:18:06 | mordred | dtantsur: in two ways ... first, we should ping mnaser to get you a vexxhost account and tobias over at citycloud will almost certainly give you one too | |
| 17:18:13 | mnaser | hi | |
| 17:18:17 | dtantsur | o/ | |
| 17:18:39 | mordred | dtantsur: but also - melvinhillsman/openlab are there to help openstack sdk authors - and setting up CI jobs for sdks against existing public cloudsis part of their remit | |
| 17:19:00 | mordred | mnaser: tl;dr - dtantsur is working on a rust sdk for openstack but only has an account on rdo cloud to test it against | |
| 17:19:03 | elmiko | mordred: that is cool (re: version scanning), are you gonna make a report or something from it? | |
| 17:19:29 | dtantsur | mordred: ah, yeah, I was thinking about openlab. a real CI job would be a big leap forward. | |
| 17:19:42 | dtantsur | currently I only have a handful of unit tests | |
| 17:19:47 | mordred | mrhillsman: ^^ I'm talking about you - and think I have another customre for you | |
| 17:19:55 | mordred | elmiko: yup- that's the idea | |
| 17:19:59 | elmiko | mordred++ | |
| 17:20:00 | mrhillsman | hey hey | |
| 17:20:03 | elmiko | very cool | |
| 17:20:17 | mordred | elmiko: someone at red hat sent me an email for advice - which has resulted in my nerd-sniping myself into writing it | |
| 17:20:25 | mrhillsman | we have space for all the SDKs :) | |
| 17:20:27 | mnaser | dtantsur: i can give you an account if you'd like, just sign up here - http://secure.vexxhost.com/billing/register.php - let me know of a client id and then i'll get it setup for ya | |
| 17:20:39 | dtantsur | mnaser: wow, awesome! | |
| 17:20:42 | mordred | elmiko: (because it should not be hard to write, but currently requires you know a bunch of internals - so step one is "fix base libs") | |
| 17:20:45 | dtantsur | mrhillsman: mine is https://github.com/dtantsur/rust-openstack | |
| 17:20:47 | dtantsur | :) | |
| 17:20:48 | mrhillsman | our hope is that you utilize openlab | |
| 17:20:59 | mrhillsman | we want to help build the sdks up | |
| 17:21:09 | mrhillsman | as our primary focus ;) | |
| 17:21:29 | elmiko | mordred: sweeet! i'm all for a good nerd-sniping ;) | |
| 17:21:41 | mrhillsman | https://github.com/theopenlab/cicd-howto | |
| 17:21:44 | elmiko | that sounded so wrong... | |
| 17:21:51 | mrhillsman | all we have to do is get the github app installed | |
| 17:21:53 | mnaser | mordred: is that a bug that things like load-balancer or object-store or placement are returning an empty list? | |
| 17:21:59 | mrhillsman | ready to make it happen right now? | |
| 17:22:03 | mordred | mnaser: it's not a bug in the script :) | |
| 17:22:10 | mrhillsman | i can work with you now to get it setup | |
| 17:22:17 | mordred | mnaser: neither load-balancer or object-store have version discovery documents | |
| 17:22:20 | mnaser | so is our service registry wrong | |
| 17:22:20 | mnaser | oh okay | |
| 17:22:37 | mnaser | http://load-balancer-ca-ymq-1.vexxhost.net/ octavia does :> | |
| 17:22:38 | mordred | mnaser: I believe johnsom may have already fixed octavia in master? | |
| 17:22:49 | mnaser | except you cant access v1, because its meant to be accessed only by neutron | |
| 17:22:52 | mnaser | so technically its v2 only | |
| 17:22:52 | mordred | mnaser: it does? awesome - lemme figure out what went wrong with it then | |
| 17:22:58 | mnaser | also wait | |
| 17:23:08 | mnaser | is it returning json wrapped in xml? | |
| 17:24:21 | mordred | mnaser: looks fine when I do it manually: | |
| 17:24:23 | mordred | get('http://load-balancer-ca-ymq-1.vexxhost.net/').content | |
| 17:24:25 | mordred | b'{"versions": [{"status": "SUPPORTED", "updated": "2014-12-11T00:00:00Z", "id": "v1"}, {"status": "CURRENT", "updated": "2017-06-22T00:00:00Z", "id": "v2.0"}]}' | |
| 17:24:33 | mnaser | mordred: ok no thats just because the browser is probably has a weird accept: clause | |
| 17:24:53 | mnaser | using accept: application/xml, it retuns <result>{json}</result> | |
| 17:24:59 | dtantsur | mrhillsman: thanks! what's your time zone? it's a bit late here | |
| 17:25:10 | johnsom | I haven't tackled the version discovery document issue yet. It dropped off my radar. | |
| 17:25:15 | mrhillsman | US Central | |
| 17:26:19 | johnsom | mordred I think you were going to open a bug with what we did wrong and/or pointer to an example done right. | |
| 17:26:35 | mordred | OH | |
| 17:26:42 | mordred | I remember what's wrong now | |
| 17:26:51 | mordred | mnaser: if you look, you will see there are no urls in that document returned | |
| 17:27:03 | mordred | johnsom: ++ I can do - I guess I should file an actual bug | |
| 17:27:32 | mrhillsman | dtantsur: i created issue in the SDK repo as well as under the OpenLab repo for tracking | |
| 17:27:35 | mnaser | https://github.com/openstack/octavia/blob/master/octavia/api/root_controller.py | |
| 17:27:38 | mnaser | looks like an easy fix :> | |
| 17:27:42 | johnsom | https://storyboard.openstack.org | |
| 17:28:01 | johnsom | Yeah, probably is, just wasn't sure what we needed. | |
| 17:28:29 | dtantsur | mrhillsman: cool! I'll read the docs and ping you agai | |
| 17:28:35 | mrhillsman | ++ | |
| 17:30:10 | mordred | mnaser, cdent: also, I do get this: keystoneauth1.exceptions.connection.ConnectFailure: Unable to establish connection to https://placement-ca-ymq-1.vexxhost.net: HTTPSConnectionPool(host='placement-ca-ymq-1.vexxhost.net', port=443): | |
| 17:30:30 | mnaser | i think its because we don't expose the placement api over our public load balancer | |
| 17:30:34 | mnaser | ..should we? | |
| 17:31:00 | cdent | it depends? | |
| 17:31:03 | mordred | well - if you have registered a public endpoint in your catalog I would - or otherwise just don't put placement into your catalog with a ublic interface | |
| 17:31:13 | cdent | yeah, what mordred says | |
| 17:31:27 | mnaser | oh actually | |
| 17:31:29 | mnaser | its exposed | |
| 17:31:33 | cdent | it does have to be in the service catalog, otherwise nova won't work, but it doesn't need to be public | |