| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2020-03-11 | |||
| 11:50:26 | openstackgerrit | Dmitry Tantsur proposed openstack/openstacksdk master: Fix microversion negotiation in some bare metal node call https://review.opendev.org/712336 | |
| 13:49:43 | openstackgerrit | Merged openstack/openstacksdk master: Set min version to test node retirement https://review.opendev.org/711958 | |
| 13:51:05 | dtantsur | mordred: when you have a minute: https://review.opendev.org/#/c/712336/ | |
| 14:15:36 | mordred | dtantsur: oy. | |
| 14:24:28 | mordred | dtantsur: microversion negotiation is hard | |
| 14:24:35 | dtantsur | it is indeed.. | |
| 14:25:20 | smcginnis | It would be great if OSC supported microversions. :P | |
| 14:33:28 | dtantsur | first, we need to agree on the default behavior.. | |
| 15:19:23 | mordred | smcginnis: it does support microversion | |
| 15:19:26 | mordred | microversions | |
| 15:24:02 | mordred | dtantsur: I agree we need to agree on the default behavior. I'd like to argue that osc's default behavior should match SDK's- which is that it should negotiate for the most recent microversion it can handle for a given command | |
| 15:24:58 | mordred | this is a little self-serving of course - since the long-term plan is to port osc on top of sdk - one would imagine that at that point osc's microversion behavior would be delegated to sdk anyway - so it would be good if porting to sdk didn't produce a behavior change | |
| 15:25:40 | smcginnis | mordred: I should have put <sarcasm/> tags on there. ;) | |
| 15:26:18 | mordred | smcginnis: :) | |
| 15:27:10 | mordred | that said - the current OSC behavior if you try something like compute_api_version: 2.latest is ridiculous | |
| 15:27:34 | mordred | but that's going to be because it just defers to novaclient and novaclient's behavior in that case is ridiculous | |
| 15:28:04 | smcginnis | I pushed back on that approach in cinder. | |
| 15:30:17 | mordred | well - the bad behavior here is that 2.latest is just an alias for "2.$whatever_the_highest_Version_hardcoded_into_novaclient_is" ... and is _not_ "2.please_negotiate_for_the_latest_available_version" | |
| 15:30:50 | mordred | so if you set 2.latest and then run opensatck server list against vexxhost, you get: clean_up ListServer: Version 2.81 is not supported by the API. Minimum is 2.1 and maximum is 2.60. (HTTP 406) (Request-ID: req-d368c068-4e37-448c-8e25-74e4aa0385a9) | |
| 15:30:58 | mordred | rather than just getting interactions at 2.60 | |
| 15:48:08 | mordred | dtantsur: ooh. I think I just had an epiphany of how to support 2.latest in sdk safely | |
| 15:48:25 | mordred | WIP patch coming | |
| 15:52:38 | dtantsur | nice! | |
| 15:52:59 | dtantsur | I think in ironic OSC plugin we treat 1 like 1.latest and support it in the correct sense.. | |
| 15:53:16 | smcginnis | Yeah, couldn't OSC do the negotiation to figure out the highest? | |
| 15:57:11 | openstackgerrit | Merged openstack/openstacksdk master: Fix microversion negotiation in some bare metal node call https://review.opendev.org/712336 | |
| 16:02:27 | mordred | smcginnis: sort of - the problem is that if you just do straight latest then a new microversion can come and break you | |
| 16:02:37 | mordred | because there is additional processing of resources returned | |
| 16:03:27 | mordred | smcginnis: which is why what you _really_ want is "negotiate for the latest the client understands" (which is the power of microversions in the first place) | |
| 16:04:33 | mordred | in practice, if we start pro-actively adding support for a service's microversion as part of adding the new whatever to the server API | |
| 16:04:43 | mordred | this should have the result people are looking for and remain safe | |
| 18:33:57 | yoctozepto | mordred: may I get your attention on https://review.opendev.org/711492 ? it seems I "fixed" devstack enough to provide what sdk requires, yet it does not accept the 'example' key config | |
| 18:39:57 | mordred | yoctozepto: looking | |
| 18:48:28 | mordred | yoctozepto: https://zuul.opendev.org/t/openstack/build/d28bf0dd977942f4ac7b6331bce2ccd8/log/controller/logs/etc/openstack/clouds.yaml#23-24 | |
| 18:48:33 | mordred | yoctozepto: that's not in the right place | |
| 18:48:47 | mordred | example should be a top-level config - a sibling of clouds | |
| 18:49:40 | mordred | yoctozepto: so it's not a property of the devstack-admin cloud | |
| 18:50:30 | yoctozepto | mordred: ack, that makes sense; just notice that 'functional' actually is and sdk docs claim that 'example' should be as well: https://opendev.org/openstack/openstacksdk/src/branch/master/doc/source/contributor/clouds.yaml | |
| 18:51:03 | mordred | one sec | |
| 18:51:14 | yoctozepto | mordred: feeling confused about the dichotomy | |
| 18:51:34 | mordred | yoctozepto: BWAHAHAHA. ok. yeah | |
| 18:51:50 | mordred | I think theres totally a bug here - let me propose a patch | |
| 18:52:08 | mordred | and ... maybe let's see if we can remove the need for a separate "example" image setting | |
| 18:52:20 | yoctozepto | mordred: I'd love that | |
| 19:05:24 | openstackgerrit | Dmitry Tantsur proposed openstack/openstacksdk stable/train: Fix microversion negotiation in some bare metal node call https://review.opendev.org/712536 | |
| 19:05:56 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Rationalize examples and functional extra config loading https://review.opendev.org/712538 | |
| 19:06:02 | mordred | yoctozepto: ^^ | |
| 19:06:21 | mordred | yoctozepto: with that, you should be able to stop writing out the example config specially | |
| 19:06:44 | mordred | yoctozepto: if that all works, I'd like to change where we're writing the functional config since it's currently just flat wrong :) | |
| 19:06:59 | mordred | and thanks - that's a leftover mess from the merge | |
| 19:08:28 | yoctozepto | mordred: quickie - the patch I mentioned modifies devstack to use this config at all, I based it off the docs and got functional working | |
| 19:08:40 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Rationalize examples and functional extra config loading https://review.opendev.org/712538 | |
| 19:08:46 | yoctozepto | mordred: you can go ahead and do it the best way you want | |
| 19:08:55 | yoctozepto | mordred: devstack won't break :-) | |
| 19:09:01 | yoctozepto | mordred: and I will adapt | |
| 19:09:04 | mordred | yoctozepto: :) | |
| 19:09:25 | mordred | yoctozepto: awesome - well, let's get it update so that the docs are right and so that we're using an exposed api inside of our own tests in sdk | |
| 19:09:50 | mordred | because we _do_ use get_extra_config in other places - so it's a real way to pass this information | |
| 19:10:25 | mordred | I don't know why we weren't using it in functional/base - nor why the examples docs were wrong even though the examples code WAS doing the right thing | |
| 19:10:26 | yoctozepto | mordred: ok, then just ditch the 'functional' difference | |
| 19:11:08 | mordred | yeah - we'll drive everything in the functional tests with the functional entry | |
| 19:11:26 | mordred | then we can fix the location of the functional entry later in the file as a followup | |
| 19:12:51 | yoctozepto | mordred: sure, let me know when you have your final change on sdk side and I'll base the devstack part accordingly | |
| 19:19:57 | mordred | yoctozepto: OH - duh | |
| 19:20:19 | mordred | yoctozepto: I just realized you're adding both functional AND example in that change | |
| 19:20:53 | mordred | yoctozepto: let me make my change the correct/simple version - no need to continue to work with the thing that doesn't exist | |
| 19:22:01 | yoctozepto | mordred: yeah, I tried to tell you that | |
| 19:22:18 | yoctozepto | mordred: also found breakage in your patch, commented | |
| 19:24:16 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Rationalize examples and functional extra config loading https://review.opendev.org/712538 | |
| 19:24:54 | mordred | yoctozepto: thanks - yes you are right with your comment :) - and yes you did - I just didn't quite follow it in my brainhole | |
| 19:25:49 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Rationalize examples and functional extra config loading https://review.opendev.org/712538 | |
| 19:25:57 | mordred | yoctozepto: I think that one should be good | |
| 19:28:50 | yoctozepto | mordred: no problem, one more comment | |
| 19:35:05 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Rationalize examples and functional extra config loading https://review.opendev.org/712538 | |
| 19:35:12 | mordred | yoctozepto: good catch | |
| 19:44:10 | yoctozepto | mordred: thanks, now +1 :-) | |
| 19:47:00 | mordred | woot! | |
| 19:47:39 | mordred | Shrews: ^^ does that have any impact we should consider on nodepool | |
| 19:48:56 | Shrews | mordred: i want to say "no" | |
| 19:49:05 | mordred | Shrews: me too | |
| 19:49:08 | yoctozepto | mordred: hahaha, sorry to inform you about another bug in there | |
| 19:49:13 | mordred | yoctozepto: yay! | |
| 19:49:28 | mordred | yoctozepto: haha | |
| 19:49:53 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Rationalize examples and functional extra config loading https://review.opendev.org/712538 | |
| 19:51:25 | yoctozepto | mordred: fingers crossed now | |
| 19:52:08 | yoctozepto | mordred: https://review.opendev.org/711492 testing devstack for overall happiness | |
| 19:54:05 | mordred | cool! fingers crossed as well | |
| 19:54:06 | yoctozepto | mordred: now that I got one of your eyes on sdk tests fixing, may I have your other eye on http://lists.openstack.org/pipermail/openstack-discuss/2020-March/013213.html ? | |
| 19:54:33 | mordred | yoctozepto: hrm. | |
| 20:27:14 | yoctozepto | mordred: will I spoil it much if I tell you that was not exactly the answer I was expecting? ;D | |
| 20:29:56 | yoctozepto | also, unfortunately, still something wrong with that patch, commented there | |
| 20:36:16 | mordred | yoctozepto: no - I was pretty sure you would not enjoy that answer - I'm mulling a better one | |
| 20:38:37 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Rationalize examples and functional extra config loading https://review.opendev.org/712538 | |
| 21:27:35 | umbSublime | heh, I was about to ask if the sdks config supported adding arbitrary sections to clouds.yaml. TIL \o/ | |
| 21:32:08 | mordred | umbSublime: :) | |
| 22:26:14 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Rationalize examples and functional extra config loading https://review.opendev.org/712538 | |
| #openstack-sdks - 2020-03-12 | |||
| 00:17:57 | openstackgerrit | Felix Yan proposed openstack/cliff master: Fix compatibility with new cmd2 https://review.opendev.org/712591 | |
| 09:54:12 | dtantsur | mordred: morning! when you have a minute, this backport https://review.opendev.org/#/c/712536/ will likely make sshnaidm a bit happier | |
| 09:55:36 | sshnaidm | dtantsur, thanks! | |
| 09:57:58 | sshnaidm | dtantsur, btw take a look https://review.opendev.org/#/c/712303/ | |