Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2019-07-09
13:49:36 ITD27M01 mordred: Hello! I am in the way to catch you! Could you please leave a comment in this PR: https://github.com/ansible/ansible/pull/51401
13:50:20 ITD27M01 mordred: If you have a second for this :)
13:52:36 mordred ITD27M01: Shrews left some comments on that about needing to add self.templare.template() call which someone else also mentioned - can you either fix that or respond that you've verified it's not needed?
13:57:02 openstackgerrit Corey Bryant proposed openstack/openstackclient master: Add Python 3 Train unit tests https://review.opendev.org/637739
13:57:20 ITD27M01 mordred: Thank you, I just wanted to make sure you knew. I am not an author and will try to contact the developer to continue working.
14:00:33 mordred ah - ok, cool. maybe I'll leave that same comment on the PR
14:00:44 mordred (and sorry - my IRC to github user mapping is incomplete)
14:02:09 Shrews looks like the latest commit addresses the template thing
14:02:51 mordred yeah.
14:02:53 Shrews but not version_added
14:03:09 mordred the way these comments get displayed is very confusing to me
14:03:15 Shrews god i hate gh
14:03:23 Shrews yes, same
14:10:35 Shrews mordred: that comment from me on that PR had me questioning my sanity b/c that call is clearly there now
14:18:41 Shrews i almost want to send my response to your response directly to GH higher ups, but i doubt they'd care
14:51:34 dtantsur folks, do we bypass major version detection when endpoint_override is used?
15:05:29 mordred dtantsur: uhm ...
15:05:51 mordred dtantsur: I know that if you provide api_version _and_ endpoint_override we completely trust you
15:05:57 dtantsur I'm seeing ironic-inspector in bifrost accessed without /v1
15:06:03 dtantsur but I don't think we provide api_version
15:06:20 mordred oh - yeah - if you give endpoint_override it should be the versioned endpoint
15:07:18 mordred endpoint_override says "use this endpoint exactly" - discovery to find a new endpoint will never happen. in some cases we might hit the discovery url to get the document so we can see microversion info
15:07:46 mordred but we won't look in to it to find a versioned endpoint if an unversioned endpoint was given
15:09:16 mordred dtantsur: unfortunately, the base system has a two-layer system for finding the url to use - the keystone catalog and then service version discovery
15:09:55 mordred thinking about it now - it seems like having 2 different possible endpoint_override parameters would actually make more complete sense
15:10:19 mordred so you could say something like "catalog_endpoint_override" or "service_endpoint_override" (but with better names)
15:11:21 mordred dtantsur: for more context - we also have to have people give both endpoint_override and api_version sometimes just to tell us what the api version is because the discovery documents are broken on their cloud
15:11:31 dtantsur huh
15:11:52 dtantsur I think I feel a bit uneasy about putting versioned endpoints to configuration files
15:12:38 mordred but - maybe we could apply the heuritstic that if you provide an endpoint_override and an api_version and the endpoint you supply does not contain "v{api_version}" that we should attempt to do a discovery lookup - and if it fails just assume you gave us a versioned endpoint and that the api version you supplied is correct
15:13:22 mordred although I think we'd need a new option for that, since it would be a behavior change
15:13:45 mordred dtantsur: I agree with all of your uneasiness
15:15:04 dtantsur the practical question for me is whether to update openstacksdk or hack endpoint_override inside ironic
15:23:13 mordred dtantsur: well - this logic is actually all in keystoneauth
15:36:04 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Return empty lists for resources if neutron doesn't exist https://review.opendev.org/669869
17:54:58 Shrews mordred: gtema: fyi, looks like the openstacksdk-functional-devstack-magnum and openstacksdk-functional-devstack-masakari check jobs have been consistently failing since https://review.opendev.org/662743
17:55:50 Shrews i was unable to find any earlier change with those failures (though it may exist)
17:59:36 mordred Shrews: oh - we should fix that then
18:00:14 mordred Shrews: looks like it's just failing the has_service Magnum check
18:00:21 Shrews i cant access the logs from this coffee shop. my vpn blocks logs.o.o for some reason
18:00:35 mordred maybe it's incidental and something changed in devstack
18:01:09 Shrews thats an odd check to fail in a magnum specific test :)
18:02:28 mordred Shrews: yah
18:02:42 mordred Shrews: also - I guess it's good that we have that check
18:02:55 Shrews ++
18:08:28 Shrews heading back to the house. thought i could work from here just to get out of the house, but much too uncomfortable here. biab
#openstack-sdks - 2019-07-10
10:23:31 openstackgerrit Bin Lu proposed openstack/python-openstackclient master: add optional argument metric to set metric for routes https://review.opendev.org/637658
14:00:42 mordred slaweq: I fixed https://review.opendev.org/#/c/669869/ yesterday - have a moment to reapply your +2 ?
14:01:46 slaweq mordred: done
14:08:48 mordred slaweq: thanks!
14:30:02 slaweq mordred: yw
15:38:42 openstackgerrit Merged openstack/openstacksdk master: Return empty lists for resources if neutron doesn't exist https://review.opendev.org/669869
17:24:43 openstackgerrit Artem Goncharov proposed openstack/openstacksdk master: Rework statistics reporting https://review.opendev.org/659841
17:35:10 openstackgerrit Artem Goncharov proposed openstack/openstacksdk master: Prepare Server to be used in cloud layer https://review.opendev.org/665105
17:41:44 openstackgerrit Artem Goncharov proposed openstack/openstacksdk master: Use resource layer for compute flavors https://review.opendev.org/650903
17:59:59 openstackgerrit Artem Goncharov proposed openstack/openstacksdk master: Repair masakari FT https://review.opendev.org/670145
#openstack-sdks - 2019-07-11
02:16:16 openstackgerrit Bin Lu proposed openstack/python-openstackclient master: add optional argument metric to set metric for routes https://review.opendev.org/637658
12:35:55 openstackgerrit Grégoire Unbekandt proposed openstack/openstacksdk master: Can't create a metadata named key, clear, delete https://review.opendev.org/670293
13:58:47 openstackgerrit Grégoire Unbekandt proposed openstack/openstacksdk master: Specify store when importing an image https://review.opendev.org/670309
15:56:38 openstackgerrit Eric Fried proposed openstack/keystoneauth master: Add a per-request global_request_id https://review.opendev.org/670341
16:00:03 edleafe Let the API-SIG Office Hour begin!
16:00:07 elmiko \o/
16:00:57 edleafe elmiko: Don't know if you got the letter from Kendall Nelson about groups planning to be in Shanghai
16:01:46 elmiko i did see that, is anyone from the sig going to be there?
16:01:56 elmiko i will not be ther
16:02:19 edleafe I was hoping to ask you and dtantsur|afk, but he's afk, apparently
16:02:27 edleafe I won't be there, either
16:05:39 elmiko i don't recall if dmitry had mentioned it
16:08:52 edleafe Well, she mentioned that unless someone says "Yes, our team will be at that PTG", the assumption is that they won't
16:09:33 edleafe So if dtantsur|afk wants to host a session there, he'll have to a) let Kendall know and b) plan on being lonely :)
16:11:12 elmiko heh, good point
16:11:42 elmiko i can send an email to dmitry just to confirm if he will be there
16:18:01 cmurphy hello, i have a question to bring to the api-sig
16:18:23 cmurphy We have a change to remove some deprecated config options related to PKI tokens https://review.opendev.org/659434 (cc vishakha)
16:18:29 cmurphy for keystone
16:18:29 openstackgerrit Merged openstack/openstacksdk master: Can't create a metadata named key, clear, delete https://review.opendev.org/670293
16:18:32 openstackgerrit Merged openstack/openstacksdk master: Repair masakari FT https://review.opendev.org/670145
16:18:34 openstackgerrit Merged openstack/openstacksdk master: Add Python 3 Train unit tests https://review.opendev.org/669449
16:19:29 cmurphy When those options are unconfigured, one of the APIs can't work and to date we've been raising a 500 to indicate the server is misconfigured
16:19:40 cmurphy With this change, there will no longer be any chance for that API to work since the server is effectively always misconfigured
16:19:51 cmurphy So the question is, do we keep raising the 500 consistent with how it worked before, or do we change the response to a 410 Gone to be a little more honest about the fact that this API just doesn't work any more
16:19:59 cmurphy keeping in mind that keystone doesn't do microversions
16:20:48 elmiko so, off the top of my head if that api is no longer supported i would expect some 4xx error as opposed to 5xx since the client is in error here
16:21:28 elmiko i'm not sure if 410 is the proper status or not, i'd have to do a little more reading
16:22:21 cmurphy okay, yeah someone else made the point that a 4XX at least indicates that retrying isn't going to do any good
16:22:48 elmiko right, imo 4xx means the client needs to do something different. which in this case means don't call those apis anymore XD
16:23:48 elmiko are you leaving the routes in for backward compatibility? like, how come you don't just return 404 for those?
16:23:58 edleafe cmurphy: Since you don't provide access to the old API behavior, it would be my preference to have keystone handle this case more gracefully. IOW, if it can do what it needs to do without those config settings, just do that
16:24:34 elmiko ++ good question edleafe
16:25:28 cmurphy elmiko: yes we wanted to leave the route there to maintain some kind of backwards compatibility even though that's not really achievable in this case
16:25:58 elmiko ack
16:26:22 cmurphy edleafe: there's no way to have keystone do the same thing without those config settings, the config setting just set a location of a CA cert file that was read and returned by that GET API
16:26:44 edleafe cmurphy: you *do* realize that mordred will hunt you down and punish you, right? :)
16:26:52 cmurphy yes >.>
16:26:54 elmiko haha
16:27:35 edleafe cmurphy: So you're making what amounts to a major version change, without leaving the old version in place. Is that correct?
16:29:18 cmurphy I don't think it's that drastic, the API was already badly designed because it depended on the operator generating this file and configuring this parameter, we're just cementing in the behavior it had if the operator failed to do that
16:30:06 cmurphy and since PKI tokens are not in use 99% of clouds will already have this behavior
16:30:56 edleafe By a "major version change" I meant one where a request that worked before can no longer work the same way.
16:31:19 edleafe If it's a case of "no one uses it anyway", then that's a practical matter

Earlier   Later