| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2017-03-07 | |||
| 12:39:58 | cdent | and no more details required? | |
| 12:40:12 | sdague | cdent: I'm fine with that. | |
| 12:41:28 | cdent | cool, thanks for passing your eyes over it | |
| 12:42:36 | sdague | so, I honestly think we probably just need a different document on best practices on writing clients. | |
| 12:43:31 | sdague | because the audience for this document isn't really going to overlap with that audience much | |
| 12:43:37 | sdague | besides like a mordred | |
| 12:56:03 | cdent | sdague: yeah, that's probably a reasonable point. I just wanted to make sure it got down somewhere for people to look at so we could make that decision | |
| 12:56:22 | cdent | having a different doc is a good idea | |
| 12:58:40 | sdague | I have to admit, playing with this blockdiag markup for the quota models doc has also given me some thoughts about effective diagramming for the microversion consumers guide | |
| 13:23:11 | cdent | sdague: that would be cool | |
| 13:52:03 | dtroyer_zz | stevemar: are you travelling today? | |
| 14:04:03 | dave-mccowan | Hello API WG. I have some questions about micro-versions and how to use them. Is now a good time? | |
| 14:08:31 | cdent | dave-mccowan: I can try to help, now, but if we can't figure it out a post to the os-dev list or coming to the meeting on thursday might help. What's up? | |
| 14:09:25 | dave-mccowan | this is for Barbican. During the Pike cycle, we'd like to make a few changes to our API, and I want to make sure to do it right. I think micro versions is the way to go. | |
| 14:10:32 | dave-mccowan | 1) we want to add a field to several of our responses, 2) we'd like to create an alias for one of our resources, 3) we'd like to start deprecating a resource that we announced we would deprecate last cycle. | |
| 14:11:25 | dave-mccowan | first... can all of those things be done as part of a microversion. next... is there a doc somewhere describing how to do a microversion? | |
| 14:12:44 | cdent | dave-mccowan: there's a couple of things that might be useful: | |
| 14:13:06 | cdent | this is the microversion spec http://specs.openstack.org/openstack/api-wg/guidelines/microversion_specification.html which describes how they are supposed to work | |
| 14:13:38 | cdent | there is a review of the guidelines for when a microversion is needed here: https://review.openstack.org/421846 | |
| 14:14:47 | cdent | and looking at the commit log for how placement implemented microversions might be useful (as it was a recent implementation) https://github.com/openstack/nova/commits/master/nova/api/openstack/placement/microversion.py | |
| 14:16:24 | cdent | dave-mccowan: to answer your questions more directly: yes, adding a field is a common microversion thing; not quite sure what you mean by alias so not sure about that; deprecating a resource is okay, but generally the concept with stability is that you keep something in the api as long as possible, even in its deprecated state | |
| 14:16:41 | cdent | with microversions you raise the minimum version if you want to fully get rid of something | |
| 14:17:11 | dave-mccowan | cdent thanks. | |
| 14:17:29 | dave-mccowan | cdent for my "alias" question, here's what i have in mind. i'd be happy to hear any advice. | |
| 14:17:32 | cdent | is that sufficient to get you started? | |
| 14:18:26 | dave-mccowan | cdent in barbican we have a resource called "container". sadly, this term has taken on a very popular non-related definition, so we'd like to rename it to something else. (bundle/group/??) | |
| 14:18:43 | cdent | :) | |
| 14:19:08 | openstackgerrit | Ian Cordasco proposed openstack/keystoneauth master: Allow new cassettes to be recorded via fixture https://review.openstack.org/442516 | |
| 14:19:09 | dave-mccowan | would making an alias for "container", and responding to both GET /container/ and GET /group/ be an OK thing to do? | |
| 14:19:34 | cdent | yes, but you could also consider a redirect | |
| 14:20:30 | sigmavirus | mordred: could you take a gander at ^ | |
| 14:20:44 | dave-mccowan | cdent ah... redirect. i like it. | |
| 14:21:09 | cdent | dave-mccowan: especially if it is just a GET the impact should be low | |
| 14:21:30 | cdent | I know that in ceilometer when they moved alarms out of ceilo to aodh they did a special kind of redirect that also works for POST | |
| 14:21:40 | cdent | 307 if I remember right | |
| 14:21:58 | cdent | or mayb 308 | |
| 14:22:32 | dave-mccowan | cdent for the deprectated question... is there a recommended response for a deprecated resource? is there something better than 404? | |
| 14:23:25 | cdent | If it is deprecated, that means it still works but will stop working at some point, so 404 is no good. There's not been a plan for how it would work other than docs. | |
| 14:23:45 | cdent | But if you have a microversion that removes the functionality then in the older microversion it works, and in the newer microversion it is 404 | |
| 14:24:19 | cdent | if you want to say in the body of the 404 response "this doesn't work in this microversion" that would be up to you | |
| 14:26:22 | dave-mccowan | cdent this will get me started. i'll put our plans in a spec and advertise it to the api-wg for review. thanks! | |
| 14:26:34 | cdent | dave-mccowan: there was some discussion at the PTG about being able to forward warn about the minimum microversion being raised, bottom of this doc: https://etherpad.openstack.org/p/ptg-architecture-workgroup | |
| 14:26:46 | cdent | dave-mccowan: awesome, looking forward to it | |
| 14:47:08 | mordred | sigmavirus: yup - looking | |
| 14:47:54 | mordred | sigmavirus: looks great! | |
| 14:48:47 | sigmavirus | mordred: have another on the way | |
| 14:48:57 | sigmavirus | (patch, not patchset) | |
| 14:48:58 | mordred | woot | |
| 14:51:46 | openstackgerrit | Ian Cordasco proposed openstack/keystoneauth master: Allow users to specify a serializer easily https://review.openstack.org/442536 | |
| 15:02:08 | openstackgerrit | Bence Romsics proposed openstack/python-openstackclient master: Make MAC address of port updatable https://review.openstack.org/442543 | |
| 15:04:55 | sigmavirus | Thanks for reviewing those mordred | |
| 15:22:04 | openstackgerrit | Monty Taylor proposed openstack/os-client-config master: Add helper factory functions for use in OpenStack services https://review.openstack.org/439120 | |
| 15:40:46 | openstackgerrit | Monty Taylor proposed openstack/os-client-config master: Don't create auth plugin until get_session https://review.openstack.org/438196 | |
| 16:02:42 | mordred | dtroyer_zz: ^^ trying to get small steps towards only creating the auth plugin on demand rather than at config reading - also https://review.openstack.org/438266 I _think_ handles both current occ and the new patch simultaneously | |
| 16:10:52 | stevemar | dtroyer_zz: yep! | |
| 16:12:39 | dtroyer_zz | where are you staying? Dinner? | |
| 16:35:35 | mordred | dtroyer_zz: also - the occ/osc source integration job definitely works, because it sure does fail on the previous version of that patch | |
| 17:10:19 | openstackgerrit | Sindhu Devale proposed openstack/python-openstackclient master: OSC Extension Show https://review.openstack.org/382023 | |
| 17:54:57 | openstackgerrit | Dean Troyer proposed openstack/python-openstackclient master: Refactor ServerTests and ServerEventTests functional test classes https://review.openstack.org/442645 | |
| 18:16:35 | mordred | dtroyer_zz, stevemar: I just got http://logs.openstack.org/66/438266/2/check/gate-osc-dsvm-functional-ubuntu-xenial/a1cca3d/testr_results.html.gz on a patch - is that a known recurring issue? | |
| 18:22:33 | ankur-gupta-f4 | Seen random volume tests failing in the past. Not sure if its a consistent recurring error | |
| 18:40:30 | openstackgerrit | Davanum Srinivas (dims) proposed openstack/golang-client master: Sanitize the package names https://review.openstack.org/441500 | |
| 18:49:35 | openstackgerrit | Merged openstack/keystoneauth master: Allow new cassettes to be recorded via fixture https://review.openstack.org/442516 | |
| 19:20:12 | openstackgerrit | Merged openstack/python-openstackclient master: Add server event list and show commands https://review.openstack.org/441775 | |
| 19:21:48 | openstackgerrit | Merged openstack/os-client-config master: Add ability to skip yaml loading https://review.openstack.org/438560 | |
| 20:06:09 | openstackgerrit | Sindhu Devale proposed openstack/python-openstackclient master: OSC Extension Show https://review.openstack.org/382023 | |
| 20:06:48 | dtroyer_zz | mordred: re the OSC functional tests, I think we've had some raciness around running out of volume space if tests run too much in parallel. | |
| 20:07:19 | mordred | dtroyer_zz: gotcha. yah - I see racy volume related tests in shade functional tests too | |
| 20:08:02 | dtroyer_zz | I upped our backing file size and that helped a bunch, can't recall actaully seeing that failure since then until now | |
| 20:09:41 | mordred | dtroyer_zz: ah - ours isn't space - we have problems with deleting volumes and state transitions - it's on my list of things to dig in to deeper | |
| 20:11:30 | dims_ | dtroyer_zz : found some cleanups for package names/imports and lined up a review | |
| 20:11:36 | dims_ | (golang-client) | |
| 20:12:03 | openstackgerrit | Dean Troyer proposed openstack/python-openstackclient master: Refactor ServerTests and ServerEventTests functional test classes https://review.openstack.org/442645 | |
| 20:12:38 | mordred | dtroyer_zz, dims_: speaking of golang-client- there was a question on that etherpad about why golang-client and why not just patches to gophercloud - the answer so far is "ask dtroyer_zz " | |
| 20:13:05 | dims_ | mordred : dtroyer_zz : y that was my comment :) | |
| 20:13:52 | mordred | dims_: :) | |
| 20:14:24 | dtroyer_zz | so, yeah, I don't have a current answer to that, except that I've always wanted to have the same sort of semantics around the Go SDK as we have in the Python libs | |
| 20:14:57 | dtroyer_zz | like wrapping the ksa bits into the "normal" route for calling REST endpoints, similar to what we did with requests | |
| 20:15:00 | dims_ | dtroyer_zz : anyone else helping with that work? | |
| 20:15:11 | dtroyer_zz | jamielennox and I talked about that a bit more in ATL | |
| 20:15:25 | mordred | dtroyer_zz: ++ | |
| 20:15:39 | dtroyer_zz | dims_: no, it's been largely idel for a while now, one reason I picked that repo to use for the CTI work | |
| 20:16:24 | dims_ | ack dtroyer_zz let's see if we can rustle up some folks to help with that | |
| 20:16:29 | dtroyer_zz | although there is one group interested in continuing to use it, they've submitted some recent reviews and I understand they want to continue, even with the other changes we need to make to it yet | |
| 20:17:57 | mordred | dims_, dtroyer_zz: also, gophercloud seems a bit rackspace-centric- and is also keystonev2 centric | |
| 20:18:04 | mordred | https://godoc.org/github.com/rackspace/gophercloud#AuthOptions being a good example of that | |
| 20:18:09 | dtroyer_zz | dims_: the priority there is Identity v3 auth plugins and such, then making the rest of the calling conventions match | |
| 20:18:21 | dtroyer_zz | mordred: yup | |
| 20:18:57 | dims_ | makes sense mordred dtroyer_zz | |
| 20:19:11 | dtroyer_zz | personally, I want the low-level SDK equivalent so I can guild go-osc on top of it | |
| 20:19:15 | mordred | dtroyer_zz: if I get off my lazy butt I'll try to help with clouds.yaml reading support or something - working on language 2 might help spur me to document the more undocumented things :) | |
| 20:19:43 | dtroyer_zz | mordred: that is always a good way to re-think things :) | |
| 20:19:45 | briancurtin | (i don’t work on or care about go whatsoever, but…) gophercloud shifted away from being a rackspace lib first and foremost into being an openstack lib that also supports rackspace | |
| 20:20:08 | dims_ | briancurtin : that's great to hear! | |
| 20:20:20 | mordred | briancurtin: yah - I noticed that focus for sure | |
| 20:20:32 | dtroyer_zz | briancurtin: cool. I haven't looked at it closely in a while, does it do Identity v3 yet? | |
| 20:20:36 | dtroyer_zz | (if you know) | |
| 20:20:57 | mordred | dtroyer_zz: it is not documented to, no | |
| 20:22:25 | briancurtin | dtroyer_zz: no clue | |
| 20:22:39 | mordred | briancurtin: my 'seems a bit rackspace-centric' is more drawn from the places where gophercloud decided to name various concepts differently in places and how it links to rackspace docs for concepts rather than openstack docs. etc | |