Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2018-03-29
16:08:14 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Add image_format for VEXXHOST profile https://review.openstack.org/557748
16:08:40 openstackgerrit Monty Taylor proposed openstack/os-client-config master: Replace guts with openstack.config https://review.openstack.org/549307
16:11:49 mordred SamYaple, Shrews, rods, slaweq: if you have a sec, I'd like to land 547159, 547488, 556089 and 555710 and then cut an sdk release
16:17:21 SamYaple mordred: 547488 is in merge conflict
16:22:47 mordred bother
16:46:29 openstackgerrit Monty Taylor proposed openstack/api-wg master: Update parameter names in microversion sdk spec https://review.openstack.org/557773
16:46:47 mordred dtantsur, elmiko, cdent, edleafe: ^^ there's one
16:47:33 elmiko mordred: wow, quick draw mcmordred!
16:48:02 dtantsur mordred: to be honest, I hate it
16:48:28 dtantsur I guess you were not in the room in Dublin when I was ranting about "micro"?
16:48:38 openstackgerrit Merged openstack/api-wg master: Add guideline on exposing microversions in SDKs https://review.openstack.org/532814
16:48:53 dtantsur tl;dr "micro" only makes any sense to 3-6 people involved in API SIG, and absolutely no sense to everyone else
16:50:18 mordred dtantsur: that's fair - but we have two completely different types of versions with completely different semantics
16:50:24 mordred dtantsur: I mean, I don't like the name either
16:50:53 dtantsur mordred: similarly to what we discussed today, I would like the major version to be taken from the "just version"
16:51:08 dtantsur it would be different if we did not include the major version in the microversion
16:51:21 dtantsur (see it already makes little sense - a major version is contained in a micro!)
16:51:47 mordred dtantsur: so - unfortunately, we have services that didn't implement microversions (yet) and include minor versions in their api versions
16:52:03 dtantsur oh
16:52:04 mordred glance 2.8, for instance
16:52:05 dtantsur my
16:52:07 dtantsur god
16:52:09 dtantsur ...
16:52:13 mordred dtantsur: welcome to my personal hell :)
16:52:27 dtantsur mordred: I started glance in rust-openstack, it's my hell as well now..
16:52:34 mordred :)
16:53:01 mordred dtantsur: so if we re-use, then people might erroneously think they can say get_image(api_version=2.8) - which they actually can't do
16:53:27 mordred dtantsur: yah
16:53:28 dtantsur mordred: can we then invent a better name for microversions?
16:53:39 mordred dtantsur: fine by me
16:53:44 mordred dtantsur: "per api call version"
16:54:14 dtantsur maybe we should call it "revision"? like, avoid "version" completely?
16:54:33 mordred I mean - it's actually "endpoint_version" and "call_version" - as one is used to find the endpoint in the keystone catalog and one is used to impact specific behavior on the endpoint found in the catalog
16:55:05 mordred but I don't think either of those words are good words
16:57:13 dtantsur still slightly better than "microversion" IMO
16:57:25 dtantsur anyway, I left a -1 for now with the goal of giving it some thought over the weekend
16:57:50 mordred dtantsur: fwiw, I've contended for a while that microversions aren't actually any different from regular versions from an end-user perspective, they're just an alternate versioning mechanism. the main difference is that the python-*client libraries all make using catalog version discovery basically impossible, so adding microversions was a way to add version discovery that could be used by the broken
16:57:52 mordred python-*client libraries
16:58:03 mordred when what we should have done is just fixed the client libraries and made them stop being broken
16:58:49 dtantsur yep, we should have..
16:59:00 dtantsur everytime I look at ironicclient I want to just rewrite it
16:59:06 mordred dtantsur: just delete it
16:59:10 mordred dtantsur: shift to openstacksdk
16:59:14 dtantsur lol
16:59:27 dtantsur well, it may be the eventual course of our actions
16:59:34 mordred dtantsur: ++
16:59:39 dtantsur but the problem is: people don't want to dedicate much time to these clients
16:59:46 dtantsur they're kind of secondary
17:00:26 mordred yah. exactly. this is why just using the sdk - or direct rest calls - is better - we don't have the collective resources to maintain all of python-*client
17:01:57 mordred dtantsur: but yeah - let's ponder the naming over the weekend and chat next week more (also, sorry I wasn't in the room in dublin)
17:05:53 JudeC Hey guys, we noticed an issue when using the python-openstackclient to view limits. If Cinder isn't enabled the limits command will fail reporting that it is unable to find an endpoint for Cinder.
17:06:27 JudeC I was considering proposing a "compute limits show" command to get around this issue.
17:07:22 JudeC Unless someone else has a better suggestion for us.
17:16:02 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Run normalize_keys on config for session codepath https://review.openstack.org/547159
17:16:03 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Remove the need for OpenStackConfig in CloudRegion https://review.openstack.org/547488
17:16:12 mordred SamYaple: thanks. fixed the merge conflict
17:16:58 edleafe mordred: yeah, nothing "micro" about them for users. They are full versions.
17:22:19 openstackgerrit Merged openstack/openstacksdk master: Add support for trunk ports and subports https://review.openstack.org/556089
17:24:20 cdent elmiko, dtantsur|afk , edleafe https://etherpad.openstack.org/p/api-sig-newsletter please and thank you
17:29:13 edleafe cdent: the microversion history stuff isn't really a guideline, thoguh
17:29:15 edleafe though
17:29:33 edleafe not even a "sort of a guideline"
17:31:06 dtroyer JudeC: the command needs to check for the presence of the volume service-type in the service catalog similar to how it already does for netowrk
17:31:10 cdent it's an appendix
17:31:16 cdent to the guidelines
17:31:20 cdent exactly what I said
17:31:35 cdent hmmm
17:34:08 cdent edleafe: I'm still inclined to say "[exist] to be published as an appendix to the guidelines" but I won't quibble. are you otherwise happy?
17:34:50 edleafe we break up our stuff into guidelines and meta info about the group. I'm saying let's make a third section, and put it there. We can fight over the name, but IMO it doesn't fit into either of the existing two groups
17:35:01 JudeC dtroyer, yeah but there is a void for showing limits if you don't have a volumes endpoint. That method for showing limits looks for the compute and and volumes endpoint without an option to do it if you only have nova without cinder.
17:35:31 edleafe cdent: yeah, I don't see it as an appendix. It's separate
17:36:02 cdent maybe I don't know what appendix means, doesn't matter
17:36:06 dtroyer JudeC: right, that is the bug, it should check for all three endpoints actually, as there are also clouds without Nova
17:36:18 cdent edleafe: otherwise okay?
17:37:05 edleafe yeah
17:37:11 cdent thanks
17:45:50 openstack Launchpad bug 1707960 in python-openstackclient "openstack limits show can't work with cinderles stack" [Undecided,New]
17:45:50 JudeC dtroyer, oh cool there is already an open bug for this, https://bugs.launchpad.net/python-openstackclient/+bug/1707960
17:46:25 JudeC I will put a patch together for this.
18:20:17 openstackgerrit Merged openstack/openstacksdk master: Add instance_ha service https://review.openstack.org/555710
20:37:33 edleafe dtantsur|afk: So with the toctree, how can I preview without pushing the patch? I'd like to not have to have Gerrit in my workflow like that
#openstack-sdks - 2018-03-30
02:41:45 openstackgerrit Merged openstack/openstacksdk master: Add image_format for VEXXHOST profile https://review.openstack.org/557748
08:12:07 openstackgerrit Merged openstack/openstacksdk master: Run normalize_keys on config for session codepath https://review.openstack.org/547159
13:23:34 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Remove the need for OpenStackConfig in CloudRegion https://review.openstack.org/547488
14:24:43 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Pass default_microversion to adapter constructor https://review.openstack.org/557935
14:24:43 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Refactor _get_version_arguments https://review.openstack.org/557934
14:27:18 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Pass default_microversion to adapter constructor https://review.openstack.org/557935
14:27:20 mordred dtantsur|afk: ^^ that should implement what we discussed yesterday
14:27:58 mordred TheJulia: ^^ that also sets a default microversion for ironic to 1.6 which is what the existing code was doing manually - but it might be good to get an ack that doing so isn't too crazy
14:28:30 TheJulia moderately cracray
14:28:36 TheJulia craycray :)
14:28:45 mordred :)
14:30:53 TheJulia that default really should be something higher, that was the reason for pinning those calls, I think
14:31:28 mordred TheJulia: well, as long as it's not too high... the patch above should be a no-op for ironic, yeah?
14:32:08 mordred TheJulia: (this is step one in providing a better story around actually detecting available microversions and doing things accordingly)
14:33:24 TheJulia the change that originally pinned it is at 1.11 if memory serves, and the code was updated to handle that if I'm remembering correctly. I think my pin accidently got knocked back a long time ago in a merge conflict and never got fixed.... Regardless ++ to better story. In the grand scheme of the universe, if the bifrost job passes, I'm happy
14:33:44 mordred TheJulia: :)
14:46:54 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Throw an error on missing support for default microversion https://review.openstack.org/557938
14:47:23 mordred TheJulia, dtantsur|afk: ^^ that should give at least a rudimentary support for blowing up early if the cloud is too old
15:37:23 TheJulia mordred: awesome

Earlier   Later