Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2019-09-25
15:46:51 mordred efried: uh...
15:46:55 Shrews mordred: so i return to my original question... how are we going to keep those fixes in sync in both places?
15:47:17 efried mordred: just skip to the last paste :)
15:47:57 Shrews i'm guessing this is a problem all external collections are going to have. just wondering if there was talk about that process
15:48:03 mordred Shrews: magic?
15:48:28 mordred Shrews: I think we should be able to make a zuul job that runs on changes to those that proposes patches to gerrit
15:49:12 mordred efried: yeah. like - that is not output we want to see - but it makes me wonder how anything would be working
15:49:12 efried mordred: the last ten lines or so of that output show the 401 for auth required for the versioned endpoint, which shouldn't be happening because a) endpoint shouldn't require auth, but even if it does, b) we're using a proper auth to set up the Connection (which can be proven by using strict_proxies=False and using the resulting proxy to do real things).
15:49:44 efried mordred: The most obvious guess is that version discovery is not using the auth I pass in.
15:49:51 efried it's using... some other auth, or no auth.
15:49:59 efried that would cause this behavior.
15:50:16 efried edit: this particular version discovery path
15:51:28 mordred yeah. something seems very strange there. I need to dig in to that - but it's lunchtime
15:51:41 efried ack
15:52:02 mordred that is definitely not behavior I expect :(
15:53:12 efried I'll open a story for this, I suppose.
15:53:21 mordred Shrews: it's a topic we definitely need to wrap our heads around
16:12:56 efried mordred: https://storyboard.openstack.org/#!/story/2006614
21:13:23 efried dtroyer: Whom can we bug to get https://review.opendev.org/#/c/664393/ merged? It's a dep of a couple of nova doc patches that would be nice to get into rc1
21:14:04 efried uhm, then again... I suppose that doesn't really help us unless osc is going to re-release for train, huh?
21:19:33 dtroyer efried: right, I was holding off on new features until I was reasonably sure we didn't need a 4.0.1. But that means 4.0.0 is stable/train
21:19:55 efried okay, never mind then
21:20:01 efried <waves hand> I was never here
21:20:14 dtroyer but that really only matters for distro packagers… everyone else should be using current anyway :)
21:21:46 efried psshhht, distros <rolls eyes>
#openstack-sdks - 2019-09-26
05:31:49 openstackgerrit Vishakha Agarwal proposed openstack/keystoneauth master: Generate pdf documentation https://review.opendev.org/682272
05:37:14 openstackgerrit Vishakha Agarwal proposed openstack/keystoneauth master: Generate pdf documentation https://review.opendev.org/682272
05:50:06 openstackgerrit Vishakha Agarwal proposed openstack/keystoneauth master: Generate pdf documentation https://review.opendev.org/682272
05:54:48 openstackgerrit pengyuesheng proposed openstack/shade master: Update the constraints url https://review.opendev.org/684929
10:51:13 mordred efried: I have reproduced your issue against vexxhost
11:28:38 openstackgerrit Monty Taylor proposed openstack/keystoneauth master: Add support for authenticated endpoint_data https://review.opendev.org/685042
11:29:45 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Use auth for discovery https://review.opendev.org/685043
11:31:18 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Use auth for discovery https://review.opendev.org/685043
11:31:33 mordred efried: ^^ those two should fix the issue
11:32:03 mordred although I think there's an improvement I want to make ...
11:38:14 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Use auth for discovery https://review.opendev.org/685043
11:38:21 openstackgerrit Monty Taylor proposed openstack/keystoneauth master: Add support for authenticated endpoint_data https://review.opendev.org/685042
11:38:41 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Send auth tokens for discovery https://review.opendev.org/685043
11:39:23 mordred efried, dtantsur, cmurphy, kmalloc: ^^ that should be ready for review now and should fix efried's issue
11:39:44 dtantsur okie, will try to get to it (release time in the ironic world)
11:40:56 mordred dtantsur: it's ... you're going to enjoy it I'm certain
11:41:08 mordred and by enjoy, I think I mean "will be sad"
11:41:11 dtantsur I've been enjoying it for a few weeks already :D
11:41:45 mordred yay for enjoyment! :)
13:42:08 efried mordred: Thanks! I'll take a look at that very soon.
13:42:32 efried mordred: were you planning to cut another Train release?
13:43:09 efried I guess that's a backportable bugfix in any case. (Is SDK on a cycle release model anyway?)
15:01:47 brtknr mordred: would you mind blessing this: https://github.com/ansible/ansible/pull/62874
15:02:28 brtknr or anyone else for that matter
15:54:32 efried mordred: Do we have a doc somewhere containing the edict that discovery endpoints should not require auth?
15:55:06 efried elmiko: perhaps that's something you would also know?
16:01:00 cdent it may be in that discovery doc that languished without merging for so long
16:01:00 elmiko API SIG office hour is now starting
16:01:08 cdent not sure if it ever did
16:01:41 elmiko efried: i feel like there /may/ be some guidance about that, but iirc it is in the languishing doc that cdent is referring to
16:01:52 elmiko at the least, i think it came up for discussion
16:02:19 elmiko i have been reviewing these open pull requests, but it has been slow going
16:05:08 elmiko efried: it /might/ be burried in here https://review.opendev.org/#/c/459405/
16:07:14 elmiko efried: this document certainly seems to imply that they should be public
16:07:17 elmiko http://specs.openstack.org/openstack/api-sig/guidelines/discoverability.html
16:08:25 efried thanks
16:09:47 efried I'm going to open a bug against nova for not doing that (for the versioned disc endpoint)
16:09:53 efried ...once this meeting is done
16:10:17 elmiko ok, i am working to try and get the rest of these version discovery docs merged. it will probably take another few weeks though
16:10:37 elmiko and they might have some warts, but hopefully we can file them down ;)
16:15:02 efried https://bugs.launchpad.net/nova/+bug/1845530 for now, will update once I find that doc...
16:15:02 openstack Launchpad bug 1845530 in OpenStack Compute (nova) "Versioned discovery endpoint should not require authentication" [Undecided,New]
16:22:47 elmiko efried: so, are we talking about just service discovery?
16:29:03 elmiko i just want to point this out, http://specs.openstack.org/openstack/api-sig/guidelines/consuming-catalog.html#discovery-algorithm
16:29:29 elmiko the "basic process" described there seems to indicate that a user /should/ authenticate to get the service endpoints
16:29:36 elmiko but i'm not sure if this is the same topic
16:30:07 elmiko oh yeah, disregard. i just re-read your bug
16:33:32 elmiko efried: found it
16:33:35 elmiko https://review.opendev.org/#/c/459710/17/guidelines/discoverability.rst
16:33:40 elmiko look at line 93
16:33:47 elmiko i'll add a comment on the bug you made
17:05:22 efried elmiko: thanks for that add, much appreciated.
17:37:24 openstackgerrit Colleen Murphy proposed openstack/keystoneauth master: Follow the PTI for docs https://review.opendev.org/685138
17:39:26 kmalloc hm... guang is not here
17:40:02 kmalloc cmurphy: https://review.opendev.org/#/c/677239/ this review is going to need to be watched. i removed my -2 on it, but this is absolutely not the fix.
17:40:10 cmurphy kmalloc: he's in -keystone
17:40:17 kmalloc ah wait. damn wrong channel
18:13:34 openstackgerrit Colleen Murphy proposed openstack/keystoneauth master: Follow the PTI for docs https://review.opendev.org/685138
18:41:09 openstackgerrit guang-yee proposed openstack/keystoneauth master: Generate pdf documentation https://review.opendev.org/682272
21:55:19 efried mordred, elmiko: FYI: https://review.opendev.org/#/q/topic:bug/1845530+(status:open+OR+status:merged)
21:55:40 efried I'm too dumb to figure out api-paste.ini
22:00:29 elmiko efried: ack, tnx
#openstack-sdks - 2019-09-27
05:06:24 openstackgerrit Shogo Saito proposed openstack/openstacksdk master: Adding basic implementation for Accelerator(Cyborg) https://review.opendev.org/679914
06:14:01 openstackgerrit Merged openstack/openstacksdk master: Make proxy honor raise_exc in REST primitives https://review.opendev.org/684433
07:28:40 openstackgerrit Vishakha Agarwal proposed openstack/keystoneauth master: Generate pdf documentation https://review.opendev.org/682272
07:46:29 openstackgerrit Shogo Saito proposed openstack/openstacksdk master: Adding basic implementation for Accelerator(Cyborg) https://review.opendev.org/679914
08:45:21 openstackgerrit Bernhard M. Wiedemann proposed openstack/keystoneauth master: Make tests pass in 2020 https://review.opendev.org/681103
10:24:32 openstackgerrit Neil Jerram proposed openstack/python-openstackclient master: Allow for router resource without 'routes' attribute https://review.opendev.org/685312
10:49:10 mordred efried: cool. I'd be happy to backport to train. need to see what cmurphy thinks about backporting the ksa patch
10:49:23 openstackgerrit Monty Taylor proposed openstack/keystoneauth master: Add support for authenticated endpoint_data https://review.opendev.org/685042
11:40:09 mordred cmurphy, kmalloc: ^^ re that - I was just thinking - instead of that approach - what if we made the discovery call do try: request(auth=False) except 401: request(auth=True) - so that we'll still work with auth-needing discovery docs, but we'll gradually go back to the general noauth case as projects stop requiring it?
11:40:37 mordred it's a little less efficient for the auth requiring case, due to 2 calls
11:51:54 efried mordred: I think that's a reasonable idea. Log a shame warning for the service to unauth their discovery endpoints. Should happen just once per client conn, so hopefully not too intrusive, nah?
11:52:26 efried hacking in new kwargs, we kinda have to support them forever-ish, which is :(
11:59:13 openstackgerrit Bernhard M. Wiedemann proposed openstack/keystoneauth master: Make tests pass in 2020 https://review.opendev.org/681103

Earlier   Later