| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2018-11-16 | |||
| 16:01:54 | johnsom | larsks I know a bunch of changes were made to API version discovery around that time in openstacksdk. I would try adding an explicit version "v2.0" to your cloud definition for the load-balancer service. | |
| 16:02:24 | larsks | johnsom: like this? load_balancer_api_version: "v2.0" | |
| 16:02:45 | johnsom | It might be v2 though. load-balancer_api_version I think | |
| 16:03:03 | johnsom | The service type has a hyphen in it | |
| 16:03:05 | larsks | I'm pretty sure that all "-" get convereted to "_" for use as keys, but I can put both in. | |
| 16:06:12 | larsks | johnsom: I've tried this: http://termbin.com/byhkv I've set the api version to "2", "2.0", "v2", and "v2.0", all seem to result in the same error. | |
| 16:10:36 | gtema | larsks: would you try adding `openstack.enable_logging(debug=True, http_debug=True)` into your test to see which requests are being sent? I'm extending SDK with own services, which also do not properly support discovery but I have not faced such behaviour so far | |
| 16:11:04 | larsks | gtema: sure, just a second. | |
| 16:13:50 | larsks | gtema: there are a bunch of "Skipping invalid version data. Missing links" in the debug output: http://termbin.com/tgff | |
| 16:20:23 | gtema | larsks: hmm. The version discovery response is certanly bad. In each version item there should be "links" attribute | |
| 16:21:46 | gtema | larsks: i guess if no response would have been sent - it would be better than that. To tell the truth - I have no clue so far. Parsing of this document is not done in SDK, but keystoneauth1 (I guess) | |
| 16:22:33 | gtema | but I would rather ask - why is this response wrong? | |
| 16:22:52 | gtema | https://wiki.openstack.org/wiki/VersionDiscovery#links | |
| 16:22:58 | larsks | As far as I know, that's just what this version of octavia spits out. | |
| 16:23:21 | larsks | johnsom: ^^^^ is that correct? | |
| 16:23:32 | gtema | sure, but I mean why is it spitting this out | |
| 16:24:28 | gtema | there might be bug in octavia (this version) or configuration. Since I do not have Octavia - I do not know for sure | |
| 16:24:45 | larsks | That I don't know. I just wish there were some way to skip the version discovery and *tell* openstacksdk what to use. | |
| 16:25:51 | johnsom | Well, the version discovery in openstacksdk changed, this is why octavia changed and added the links in a later version. Still, openstacksdk can use Octavia, it's how the CLI and horizon dashboard work. We just have some configuration issue in openstacksdk that is causing it to not be happy for some reason. | |
| 16:26:47 | johnsom | Yeah, there definitely is a way to tell the SDK to use a specific version, but you tried the way I know. Sadly, I haven't spent much time with openstacksdk in over a year. | |
| 16:26:56 | gtema | there is a way, but just temporarily - `openstacksdk:openstack.load_balancer.v2._proxy.Proxy.skip_discovery = True` | |
| 16:27:08 | gtema | but you will not get it landed in SDK | |
| 16:28:33 | gtema | johnsom: the problem is that octavia in this case does not return a valid version discovery doc, so I might try to process this correctly. But the root should be fixed | |
| 16:29:05 | johnsom | gtema It returned a valid discovery doc for the Queens timeframe | |
| 16:29:20 | johnsom | gtema He is running Queens | |
| 16:29:55 | gtema | are you sure links was not mandatory field? I am using Mitaka and services return this attribute | |
| 16:30:09 | johnsom | Since the CLI is working, the SDK works against the queens Octavia API | |
| 16:30:26 | johnsom | Yep, it was not required until later | |
| 16:30:44 | gtema | CLI is working because octaviaclient is not using SDK (I guess) | |
| 16:31:41 | gtema | yesm octaviaclient is not using SDK as of now | |
| 16:32:25 | johnsom | Ah, it was just octavia-dashboard that used the SDK at that time | |
| 16:32:59 | larsks | Bleargh. | |
| 16:33:18 | larsks | Okay, I'm going to punt on this right now and just right my playbook using the openstack cli rather than os_loadbalancer for now. | |
| 16:33:38 | johnsom | We just need monty or someone that knows about that version discovery change to tell us what the configuration magic is for the SDK | |
| 16:34:04 | larsks | I'm going to spend a little time stepping through the code with pdb and see if maybe it's obvious... | |
| 16:34:30 | gtema | I know that you can disable version discovery per service (from the code, not configurable), but I am quite sure for Octavia it will not be switched off that easy | |
| 16:34:49 | gtema | `openstack.load_balancer.v2._proxy.Proxy.skip_discovery = True` | |
| 16:35:45 | larsks | gtema: isn't that at the wrong place? The problem right now is that it doesn't know to use the load_balancer.v2 proxy because of the version discovery issue... | |
| 16:36:48 | gtema | larsks: if the discovery is skipped - it should not even try to get it and return a "proper" proxy | |
| 16:37:18 | johnsom | larsks Try one more thing, looking at the code it bypasses that if both the version and the endpoint of overriden in the config | |
| 16:38:15 | larsks | gtema: setting openstack.load_balancer.v2._proxy.Proxy.skip_discovery=True doesn't seem to impact the issue. | |
| 16:38:18 | larsks | johnsom: let me try that... | |
| 16:38:33 | johnsom | Since you are using the old port way of doing endpoints, append "/v2" to what you have in keystone and try that as an override | |
| 16:38:39 | larsks | Actually, what's the syntax for that...load_balanacer_endpoint? | |
| 16:38:50 | larsks | ...but spelled correctly. | |
| 16:39:33 | johnsom | my "openstack endpoints" gives http://10.21.21.53/load-balancer and uses "http://10.21.21.53/load-balancer/v2" for the sdk endpoint. | |
| 16:40:26 | larsks | johnsom: success! | |
| 16:40:36 | gtema | great | |
| 16:41:06 | larsks | (although apparently octavia lives at /v2.0) | |
| 16:41:21 | johnsom | Yeah, so something might have changed in the newer sdk that broke the old funky stuff that was happening in queens. Monty did a ton of work here to fix the SDK version discovery stuff. | |
| 16:41:55 | gtema | api-ref says /v2.0 would be an alias to /v2 | |
| 16:42:16 | johnsom | Yeah, /v2.0 is good and will work | |
| 16:42:40 | larsks | Aaaaand now os_loadbalancer is happy, and I can rewind my stack back a couple of hours and continue :). Thanks johnsom , gtema ! | |
| 16:43:26 | gtema | larsks: congrats for "funny" friday ;-) | |
| 17:21:18 | gtema | mordred: have you arrived already? | |
| 17:24:50 | gtema | mordred or anyone else who might know: all latest pathes fail in `bifrost-integration-tinyipa-ubuntu-xenial` with `Unable to establish connection to http://127.0.0.1:5050: HTTPConnectionPool(host='127.0.0.1', port=5050): Max retries exceeded with url` | |
| 17:24:59 | gtema | I was not able to figure out the reason yet | |
| 18:18:47 | frickler | gtema: TheJulia: ^^ that looks like something in ironic may have broken http://logs.openstack.org/98/618198/1/check/bifrost-integration-tinyipa-ubuntu-xenial/14c183a/logs/ironic-inspector.log | |
| #openstack-sdks - 2018-11-17 | |||
| 00:58:18 | openstackgerrit | Merged openstack/keystoneauth master: Add py36 tox environment https://review.openstack.org/615845 | |
| 02:52:09 | openstackgerrit | zhouxinyong proposed openstack/keystoneauth master: Replacing the HTTP protocal with HTTPS in using-sessions.rst. https://review.openstack.org/617811 | |
| 03:45:33 | openstackgerrit | Yang JianFeng proposed openstack/openstacksdk master: Add support for octavia's resuorces quota https://review.openstack.org/590193 | |
| 05:45:08 | openstackgerrit | Yang JianFeng proposed openstack/openstacksdk master: Add support for octavia's resuorces quota https://review.openstack.org/590193 | |
| 13:24:30 | openstackgerrit | Lars Kellogg-Stedman proposed openstack/openstacksdk master: syntax errors and undefined exceptions in service_description.py https://review.openstack.org/618519 | |
| 16:53:06 | isears | I'm running python-openstackclient v3.17.0 and I don't seem to have access to the 'stack' sub commands anymore (e.g. openstack stack list). Was support for those commands removed recently or is my install somehow broken? | |
| 16:56:36 | isears | lol nvm I'm an idiot. Had to pip install python-heatclient. | |
| 20:13:41 | openstackgerrit | Merged openstack/openstacksdk master: Advancing the protocal of the website to HTTPS in compute.rst. https://review.openstack.org/617762 | |
| #openstack-sdks - 2018-11-18 | |||
| 06:06:44 | openstackgerrit | Yang JianFeng proposed openstack/openstacksdk master: Add support for octavia's resuorces quota https://review.openstack.org/590193 | |
| 07:17:01 | openstackgerrit | Yang JianFeng proposed openstack/openstacksdk master: Add support for octavia's resuorces quota https://review.openstack.org/590193 | |
| 19:56:40 | openstackgerrit | Merged openstack/keystoneauth master: Replacing the HTTP protocal with HTTPS in using-sessions.rst. https://review.openstack.org/617811 | |
| #openstack-sdks - 2018-11-19 | |||
| 07:22:12 | openstackgerrit | wangxiyuan proposed openstack/openstacksdk master: Add missing properties for role https://review.openstack.org/614705 | |
| 07:33:55 | openstackgerrit | zhufl proposed openstack/openstacksdk master: Add missing seperator between words https://review.openstack.org/618704 | |
| 08:50:50 | openstackgerrit | wangxiyuan proposed openstack/openstacksdk master: Add missing properties for role https://review.openstack.org/614705 | |
| 10:55:36 | dtantsur | frickler, mordred, https://review.openstack.org/618726 will hopefull fix it | |
| 10:55:47 | dtantsur | sorry :( | |
| 10:56:52 | dtantsur | gtema: ^^ | |
| 10:58:13 | gtema | dtantsur: thanks | |
| 12:00:09 | dtantsur | the fix has merged, feel free to recheck stuff | |
| 16:37:46 | openstackgerrit | Ed Leafe proposed openstack/api-sig master: Add idempotent DELETE guidance https://review.openstack.org/616610 | |
| 21:27:40 | cfriesen | dtroyer: I was just trying to get a VNC console URL via the "openstack" client, and I found it very unintuitive that I needed to run "openstack console url show" rather than something like "openstack server console show". The console URL is arguably an attribute of the server, not a standalone thing. | |
| 21:30:58 | dtroyer | cfriesen: where were you 5 years ago? :) | |
| 21:32:11 | dtroyer | I agree the resource naming is sub-optimal, that was in the early batches of things I constructed and the rules were not yet fully-formed. | |
| #openstack-sdks - 2018-11-20 | |||
| 06:15:29 | openstackgerrit | Filippo Inzaghi proposed openstack/openstackclient master: removing older python version 3.4 from setup.cfg https://review.openstack.org/606361 | |
| 09:01:36 | openstackgerrit | wangxiyuan proposed openstack/openstacksdk master: Add missing properties for role https://review.openstack.org/614705 | |
| 12:01:27 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Refactor tag support https://review.openstack.org/618993 | |
| 14:27:21 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: increase timeout not to fail functional tests https://review.openstack.org/619008 | |
| 14:31:50 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: increase timeout in heat functional test https://review.openstack.org/619008 | |
| #openstack-sdks - 2018-11-21 | |||
| 07:07:06 | openstackgerrit | Shilpa Devharakar proposed openstack/openstacksdk master: Add functional tests for masakari https://review.openstack.org/615744 | |
| 10:54:13 | Miouge | I’m struggling to set the container X-Versions-Location on a container with the CLI. It looks like “openstack container set foo —property versions_location=bar” should work, but it gets added as “X-Container-Meta-versions_location: archive” (https://github.com/openstack/openstacksdk/blob/master/openstack/object_store/v1/container.py#L23) any ideas? | |
| 10:59:10 | Miouge | On the other hand “swift post foo --header 'X-Versions-Location:bar’” works Oo | |
| 12:40:45 | larsks | Hey folks. I'm trying to submit a review (https://review.openstack.org/#/c/618519/), but it keeps failing CI due to a timeout in openstack-tox-py27. Is this a known problem? It's been rechecked several time and keeps getting different sets of failures that seem unrelated to the change itself. | |
| 12:41:39 | gtema | larsks: I am working on the change to fix timeouts: https://review.openstack.org/#/c/619008/ So far you can set it is dependency to your change | |
| 12:41:50 | larsks | gtema: thanks, I'll give that a shot. | |
| 12:42:17 | gtema | it will hopefully land soon. I am currently reworking to create a nicer/configurable solution | |
| 12:43:21 | openstackgerrit | Lars Kellogg-Stedman proposed openstack/openstacksdk master: syntax errors and undefined exceptions in service_description.py https://review.openstack.org/618519 | |
| 13:03:38 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Make timeouts in functional tests configurable https://review.openstack.org/619008 | |
| 13:56:37 | ShilpaSD | mordred:Hi, addressed review comments on https://review.openstack.org/#/c/615744/2/, kindly further review. | |
| 14:46:22 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Make timeouts in functional tests configurable https://review.openstack.org/619008 | |
| 16:33:32 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Make timeouts in functional tests configurable https://review.openstack.org/619008 | |
| 17:56:37 | timburke | Miouge: iirc, the openstack cli treats --property similar to how the swift cli treats --meta; that is, it stores the key-value pairs as user metadata. i don't know that unified client has a way to specify arbitrary headers... | |
| 18:07:37 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Make timeouts in functional tests configurable https://review.openstack.org/619008 | |