| openstackgerrit | Tsuyoshi Nagata proposed openstack/nova master: openstack server add volume fails over 26vols https://review.openstack.org/567472 | 00:09 |
|---|---|---|
| openstackgerrit | Chris Dent proposed openstack/nova master: [placement] Do not import oslo_service for log_options https://review.openstack.org/572231 | 00:14 |
| *** cdent has quit IRC | 00:20 | |
| *** tetsuro has joined #openstack-placement | 00:42 | |
| openstackgerrit | tianhui proposed openstack/nova master: Fix bug for hypervisors https://review.openstack.org/572063 | 01:31 |
| *** cdent has joined #openstack-placement | 01:56 | |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Fix wrong commands in launch-from-volume doc https://review.openstack.org/572241 | 02:16 |
| *** cdent has quit IRC | 02:27 | |
| openstackgerrit | Zhenyu Zheng proposed openstack/nova master: Use ThreadPoolExecutor for max_concurrent_live_migrations https://review.openstack.org/563505 | 02:48 |
| *** alex_xu has quit IRC | 03:00 | |
| *** alex_xu has joined #openstack-placement | 03:01 | |
| openstackgerrit | Bhagyashri Shewale proposed openstack/nova master: libvirt: Don't report DISK_GB if sharing https://review.openstack.org/560459 | 04:47 |
| openstackgerrit | Chris Dent proposed openstack/nova master: [placement] Do not import oslo_service for log_options https://review.openstack.org/572231 | 05:10 |
| openstackgerrit | Merged openstack/nova stable/ocata: libvirt: Report the virtual size of RAW disks https://review.openstack.org/568382 | 05:38 |
| openstackgerrit | Zhenyu Zheng proposed openstack/nova master: WIP abort live migration in queue. https://review.openstack.org/568542 | 06:30 |
| *** tssurya has joined #openstack-placement | 06:45 | |
| openstackgerrit | tianhui proposed openstack/nova master: Fix bug to doc https://review.openstack.org/572268 | 06:47 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Microversion 2.63 - Use new format policy in server group https://review.openstack.org/567534 | 07:11 |
| openstackgerrit | jichenjc proposed openstack/nova master: z/VM Driver: Initial change set of z/VM driver https://review.openstack.org/523387 | 07:12 |
| openstackgerrit | jichenjc proposed openstack/nova master: z/VM Driver: Spawn and destroy function of z/VM driver https://review.openstack.org/527658 | 07:12 |
| openstackgerrit | jichenjc proposed openstack/nova master: z/VM Driver: add snapshot function https://review.openstack.org/534240 | 07:12 |
| openstackgerrit | jichenjc proposed openstack/nova master: z/VM Driver: add power actions https://review.openstack.org/543340 | 07:12 |
| openstackgerrit | jichenjc proposed openstack/nova master: z/VM Driver: add get console output https://review.openstack.org/543344 | 07:12 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Fix wrong commands in launch-from-volume doc https://review.openstack.org/572241 | 07:20 |
| openstackgerrit | Chris Dent proposed openstack/nova master: Re-base placement object unit tests on NoDBTestCase https://review.openstack.org/572279 | 07:28 |
| openstackgerrit | Chris Dent proposed openstack/nova master: Move db using provider unit tests to functional https://review.openstack.org/572280 | 07:28 |
| openstackgerrit | Chris Dent proposed openstack/nova master: Move db using allocation unit tests to functional https://review.openstack.org/572281 | 07:28 |
| openstackgerrit | Chris Dent proposed openstack/nova master: Remove nova dependencies from test_resource_provider https://review.openstack.org/572282 | 07:28 |
| openstackgerrit | Naichuan Sun proposed openstack/nova master: xenapi(N-R-P): Add API to support vgpu resource provider create https://review.openstack.org/520313 | 07:56 |
| *** ttsiouts has joined #openstack-placement | 08:09 | |
| *** ttsiouts has quit IRC | 08:22 | |
| *** ttsiouts has joined #openstack-placement | 08:31 | |
| *** ttsiouts has quit IRC | 08:32 | |
| openstackgerrit | jichenjc proposed openstack/nova master: Enhance api-guide general info some updates https://review.openstack.org/561773 | 08:34 |
| *** ttsiouts has joined #openstack-placement | 08:35 | |
| *** e0ne has joined #openstack-placement | 08:37 | |
| openstackgerrit | jichenjc proposed openstack/nova master: Trivial: let internal use only func has _ prefix https://review.openstack.org/572304 | 08:37 |
| *** bhagyashri_s has joined #openstack-placement | 08:38 | |
| openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Fix some wrong urls in doc https://review.openstack.org/571147 | 08:39 |
| openstackgerrit | jichenjc proposed openstack/nova master: Enhance api-guide general info some updates https://review.openstack.org/561773 | 08:54 |
| *** ttsiouts has quit IRC | 09:10 | |
| *** takashin has left #openstack-placement | 09:30 | |
| openstackgerrit | Lee Yarwood proposed openstack/nova stable/queens: Allow cinderv2 endpoints within the request context catalog https://review.openstack.org/572213 | 09:41 |
| *** ttsiouts has joined #openstack-placement | 09:47 | |
| *** ttsiouts has quit IRC | 09:52 | |
| openstackgerrit | Stephen Finucane proposed openstack/nova-specs master: Add 'numa-aware-vswitches' spec https://review.openstack.org/541290 | 09:56 |
| *** ttsiouts has joined #openstack-placement | 10:04 | |
| openstackgerrit | Theodoros Tsioutsias proposed openstack/nova-specs master: Add PENDING vm state https://review.openstack.org/554212 | 10:08 |
| openstackgerrit | Theodoros Tsioutsias proposed openstack/nova-specs master: Enable rebuild for instances in cell0 https://review.openstack.org/554218 | 10:27 |
| openstackgerrit | Brianna Poulos proposed openstack/nova master: Implement certificate_utils https://review.openstack.org/479949 | 10:37 |
| openstackgerrit | Brianna Poulos proposed openstack/nova master: Plumb trusted_certs through libvirt driver image paths https://review.openstack.org/561262 | 10:37 |
| openstackgerrit | Brianna Poulos proposed openstack/nova master: Add trusted_image_certificates to REST API https://review.openstack.org/486204 | 10:37 |
| openstackgerrit | Brianna Poulos proposed openstack/nova master: Add notification support for trusted_certs https://review.openstack.org/563269 | 10:37 |
| openstackgerrit | Brianna Poulos proposed openstack/nova master: Add certificate validation docs https://review.openstack.org/560158 | 10:37 |
| openstackgerrit | Brianna Poulos proposed openstack/nova master: Plumb trusted_certs through libvirt driver image paths https://review.openstack.org/561262 | 10:42 |
| openstackgerrit | Brianna Poulos proposed openstack/nova master: Add trusted_image_certificates to REST API https://review.openstack.org/486204 | 10:42 |
| openstackgerrit | Brianna Poulos proposed openstack/nova master: Add notification support for trusted_certs https://review.openstack.org/563269 | 10:42 |
| openstackgerrit | Brianna Poulos proposed openstack/nova master: Add certificate validation docs https://review.openstack.org/560158 | 10:42 |
| openstackgerrit | Lee Yarwood proposed openstack/nova stable/pike: libvirt: slow live-migration to ensure network is ready https://review.openstack.org/559032 | 10:46 |
| *** ttsiouts has quit IRC | 10:55 | |
| openstackgerrit | Lee Yarwood proposed openstack/nova stable/ocata: libvirt: slow live-migration to ensure network is ready https://review.openstack.org/559034 | 11:04 |
| *** bhagyashri_s has quit IRC | 11:16 | |
| *** bhagyashri_s has joined #openstack-placement | 11:17 | |
| *** edmondsw has joined #openstack-placement | 11:19 | |
| openstackgerrit | Chen Hanxiao proposed openstack/nova master: sync_guest_time: use the proper errno https://review.openstack.org/572346 | 11:26 |
| *** tetsuro has quit IRC | 11:26 | |
| *** ttsiouts has joined #openstack-placement | 11:37 | |
| *** finucannot is now known as stephenfin | 12:44 | |
| *** mriedem has joined #openstack-placement | 12:46 | |
| *** ttsiouts has quit IRC | 12:48 | |
| *** ttsiouts has joined #openstack-placement | 12:49 | |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Transform instance.exists notification https://review.openstack.org/403660 | 12:51 |
| openstackgerrit | Jay Pipes proposed openstack/nova-specs master: Standardize CPU resource tracking https://review.openstack.org/555081 | 13:00 |
| *** ttsiouts has quit IRC | 13:45 | |
| *** ttsiouts has joined #openstack-placement | 13:58 | |
| openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: Add 'numa-aware-vswitches' spec https://review.openstack.org/541290 | 14:09 |
| openstackgerrit | Naichuan Sun proposed openstack/nova master: xenapi(N-R-P): Add API to support vgpu resource provider create https://review.openstack.org/520313 | 14:11 |
| openstackgerrit | Naichuan Sun proposed openstack/nova master: xenapi(N-R-P): Add API to support vgpu resource provider create https://review.openstack.org/520313 | 14:23 |
| openstackgerrit | Naichuan Sun proposed openstack/nova master: xenapi(N-R-P):Get vgpu info from `allocations` https://review.openstack.org/521717 | 14:26 |
| openstackgerrit | Naichuan Sun proposed openstack/nova master: xenapi(N-R-P): support compute node resource provider update https://review.openstack.org/521041 | 14:27 |
| *** mriedem has quit IRC | 14:53 | |
| *** mriedem has joined #openstack-placement | 14:53 | |
| efried | jaypipes: I've been assuming you're going to be the one to host the hangout, yah? | 14:53 |
| * gibi will try to call in as well | 14:55 | |
| *** cdent has joined #openstack-placement | 14:58 | |
| jaypipes | efried, gibi, cdent, edleafe, dansmith, mriedem, melwitt, bauzas, alex_xu: hangout for the "migration API" in placement: https://hangouts.google.com/call/N7Pb0gg01m30ep7hHzCNAAEE | 15:01 |
| bauzas | now ? | 15:01 |
| bauzas | jaypipes: now ? | 15:01 |
| cdent | yes now | 15:01 |
| jaypipes | cdent: coming back? | 15:03 |
| cdent | yeah, my sound is busted, so gotta fix that | 15:03 |
| dansmith | cdent: nice. | 15:08 |
| openstackgerrit | Surya Seetharaman proposed openstack/nova-specs master: Handling a down cell https://review.openstack.org/557369 | 15:08 |
| openstackgerrit | Theodoros Tsioutsias proposed openstack/nova-specs master: Add PENDING vm state https://review.openstack.org/554212 | 15:19 |
| jaypipes | cdent: you speaking? can't hear you... | 15:22 |
| cdent | jaypipes: nope, not yet, everybody else keeps saying what I would say | 15:22 |
| jaypipes | :) | 15:22 |
| edleafe | /win 16 | 15:24 |
| *** ttsiouts has quit IRC | 15:39 | |
| *** ttsiouts has joined #openstack-placement | 15:44 | |
| mriedem | if you can't change the parent provider uuid on an RP, we should probably call that out in the api reference https://developer.openstack.org/api-ref/placement/#update-resource-provider | 15:50 |
| mriedem | since it's not clear to me | 15:50 |
| *** e0ne has quit IRC | 15:52 | |
| mriedem | btw, i don't see anything in the handler code that prevents you from updating the parent_provider_uuid on an existing rp | 15:53 |
| *** tssurya has quit IRC | 15:57 | |
| edleafe | mriedem: yeah, I just looked too. Looks like can reparent an RP with PUT | 15:57 |
| openstackgerrit | Merged openstack/nova-specs master: Follow the new PTI for document build https://review.openstack.org/551802 | 16:13 |
| *** ttsiouts has quit IRC | 16:15 | |
| efried | edleafe, mriedem: https://github.com/openstack/nova/blob/master/nova/tests/functional/api/openstack/placement/gabbits/resource-provider.yaml#L588 | 16:23 |
| mriedem | efried: ok that's way down in the db update code | 16:26 |
| mriedem | would have been much more easy to find if it were controlled in the API | 16:26 |
| efried | I didn't look, but iirc it's an arbitrary check, not there for any technical reason. | 16:27 |
| efried | i.e. something we could remove easily (with a microversion) | 16:27 |
| mriedem | http://git.openstack.org/cgit/openstack/nova/tree/nova/api/openstack/placement/objects/resource_provider.py#n1026 | 16:27 |
| mriedem | sure, just need to call it out in the api-ref as a limitation i think; i'll push a patch | 16:27 |
| efried | ight | 16:28 |
| openstackgerrit | Merged openstack/nova master: Fix bug for hypervisors https://review.openstack.org/572063 | 16:35 |
| openstackgerrit | Merged openstack/nova master: Fix bug to doc https://review.openstack.org/572268 | 16:36 |
| gibi | efried: as soon as you allow reparenting you allow moving RPs inbetween trees and that might break existing allocations that requires to have the provider within the same tree | 16:36 |
| cdent | jaypipes: this is the start of a series which cleans out db stuff from the test_resource_provider unit test. https://review.openstack.org/#/c/572279/ That ought to lessen some of the test changes you needed to make in the consumer gen tree. | 16:39 |
| edleafe | gibi: what about allowing reparenting within the same tree? | 16:40 |
| efried | gibi: What about existing allocations requires them to be in the same tree? | 16:40 |
| efried | That was weird, I was like, "did I press Enter? I didn't *think* I pressed Enter. Enter. Oh. Hi Ed." | 16:41 |
| jaypipes | cdent: ooh, cool, ty sir! | 16:41 |
| gibi | edleafe, efried: yeah, if we limit that the new parent can only be in the same tree as the old that might help | 16:43 |
| efried | gibi: I'm asking why it's a limitation that you can't move an allocation from one tree to another. | 16:43 |
| efried | I mean, makes sense in the nova universe (for now). But purely from a placement standpoint... why? | 16:44 |
| gibi | efried: because the server create request which resulted in the allocations assumes that the resources are from the same compute tree | 16:44 |
| gibi | efried: ahh, gotcha | 16:44 |
| gibi | efried: so nova needs to make sure that the reparenting of a server allocation does not break the assumptions of nova having resources from a single tree | 16:45 |
| efried | gibi: I don't think we're considering removing the restriction; just kind of noticing that it's there and reevaluating it. The path of least resistance will be to leave it alone until we hear some use case that needs to do it. | 16:47 |
| gibi | still I feel that when the placement was called to provide allocation candidates placement made the same assumption, i.e. unnumbered group expected to be fulfilled from the same tree | 16:47 |
| gibi | efried: OK, that sound enough for me now :) | 16:47 |
| *** e0ne has joined #openstack-placement | 16:48 | |
| *** e0ne_ has joined #openstack-placement | 16:50 | |
| openstackgerrit | Merged openstack/nova stable/queens: Allow cinderv2 endpoints within the request context catalog https://review.openstack.org/572213 | 16:52 |
| *** e0ne has quit IRC | 16:54 | |
| openstackgerrit | Merged openstack/nova-specs master: Add 'numa-aware-vswitches' spec https://review.openstack.org/541290 | 16:56 |
| *** cdent has quit IRC | 16:59 | |
| *** cdent_ has joined #openstack-placement | 16:59 | |
| *** e0ne_ has quit IRC | 17:08 | |
| edleafe | efried: I definitely think we should remove (or at least relax) that restriction | 17:11 |
| efried | edleafe: I don't disagree, but until we have a compelling use case for it, that's just work on already-full plates. | 17:11 |
| efried | needs to be a microversion regardless - we can't ex-post-facto it - so... | 17:12 |
| * cdent_ suddenly hungry | 17:12 | |
| *** cdent_ is now known as cdent | 17:13 | |
| edleafe | efried: well, the use case discussed in the hangout is a potential reality. IAC, I'm not saying do it now, but just keep in mind that it's certainly possible | 17:13 |
| *** e0ne has joined #openstack-placement | 17:14 | |
| efried | nod | 17:20 |
| *** tssurya has joined #openstack-placement | 17:22 | |
| openstackgerrit | Chris Friesen proposed openstack/nova-specs master: Add support for emulated virtual TPM https://review.openstack.org/571111 | 17:31 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: api-ref: mention that you can't re-parent a resource provider https://review.openstack.org/572501 | 17:45 |
| openstackgerrit | Dan Smith proposed openstack/nova master: Change consecutive build failure limit to a weigher https://review.openstack.org/572195 | 18:18 |
| *** edmondsw has quit IRC | 18:41 | |
| *** edmondsw has joined #openstack-placement | 19:07 | |
| *** edmondsw_ has joined #openstack-placement | 19:12 | |
| *** edmondsw has quit IRC | 19:15 | |
| openstackgerrit | Dan Smith proposed openstack/nova master: Change consecutive build failure limit to a weigher https://review.openstack.org/572195 | 19:38 |
| jaypipes | efried: still working on tetsuro's patches... | 19:42 |
| jaypipes | efried: just letting you know. | 19:42 |
| *** e0ne has quit IRC | 19:54 | |
| efried | jaypipes: ack, thx. Still working on that spec :P | 20:09 |
| *** e0ne has joined #openstack-placement | 20:10 | |
| *** edmondsw_ is now known as edmondsw | 20:10 | |
| jaypipes | efried: thx | 20:11 |
| *** e0ne has quit IRC | 20:14 | |
| *** e0ne has joined #openstack-placement | 20:21 | |
| *** e0ne has quit IRC | 20:25 | |
| efried | jaypipes: how is the virt driver getting allocation data? Is he allowed to query placement, or does update_provider_tree get a new arg? | 20:25 |
| jaypipes | efried: see my review on https://review.openstack.org/#/c/567113/. I have modified that patch locally with tested code from my review comments. I'm happy to rebase that patch series and fix up what I believe are mistakes in it to get it going. Please check it out. | 20:35 |
| efried | ... | 20:35 |
| jaypipes | efried: good question. :) | 20:36 |
| jaypipes | efried: we could provide them as part of the ProviderTree I guess.. | 20:37 |
| jaypipes | efried: though we're already getting long in the tooth for that object, IMHO. | 20:37 |
| efried | jaypipes: Enhancing ProviderTree or a separate arg to upt, either would wfm, and I think that's better than upt doing placement GETs. | 20:39 |
| efried | jaypipes: separate arg, methinks. ProviderTree is organized by provider. We want the allocs organized by consumer (even though the allocs we want are the ones related to the providers in the tree) | 20:41 |
| jaypipes | efried: ack, makes sense to me. | 20:43 |
| efried | jaypipes: Without a new query, the alloc info is going to be kind of tricky/expensive to get. We need to GET /resource_providers/{rp_uuid}/allocations for all rps in the local tree (not sharing providers), and *then* create a set of all consumers from that so we can GET /allocations/{consumer_uuid} for each. | 20:44 |
| jaypipes | efried: we really only need to do the GET /resource_providers/{uuid}/allocations I think. | 20:47 |
| jaypipes | efried: and pass a dict of those (keyed by rp_uuid) to the upt. | 20:47 |
| efried | jaypipes: We would have to do that for all associated providers, including sharing, else we would be missing the sharing providers' allocations. BUT that means we're also getting sharing provider allocations against consumers on other hosts. | 20:48 |
| efried | hence the two-step. | 20:48 |
| jaypipes | efried: yeah, eww. | 20:49 |
| efried | jaypipes: Re https://review.openstack.org/#/c/567113/ -- it's not clear to me whether your reworked sql actually fixes a bug or is just more efficient. If there's an actual problem with forbidden traits, we should have a test case to demonstrate it. | 20:55 |
| jaypipes | efried: we do have test cases that hit it. test_all_local(), after tetsuro's patch links up get_trees_matching_all() into AllocationCandidates.get_by_requests() actually tickles the SQL code there. Tetsuro just changed the INNER JOIN to a LEFT JOIN thinking that was appropriate, but it isn't (as I tried to explain in the review there). | 21:04 |
| efried | jaypipes: Okay, so if it's not functionally wrong, for the sake of expediency, could we do the sql rejigger in a fup? | 21:05 |
| efried | jaypipes: and the other nits | 21:05 |
| jaypipes | efried: you mean all the changes that aren't relevant to the actual patch? | 21:05 |
| efried | jaypipes: I didn't look too closely; if that's all they are, then again, would it be possible to let 'em slide for the sake of getting this merged today? | 21:06 |
| efried | nits like doc rewords, param switch on _get_usages_by_provider... | 21:07 |
| jaypipes | efried: the "param switch" changed the behaviour of the method entirely, and for no reason. | 21:08 |
| jaypipes | efried: I might be willing to "let this slide", but I need to check the following patches first, ok? I am currently reviewing them. | 21:08 |
| efried | jaypipes: Roger that. | 21:09 |
| efried | I'm just worried about finding someone competent to review a massive sqla block. If we get it merged and functionally working, then it's okay if that takes a few days. | 21:10 |
| *** tssurya has quit IRC | 21:10 | |
| jaypipes | efried: and I am worried that s/join/outerjoin/ is really just a kludge to fix something further down this patch stack. | 21:11 |
| efried | ack | 21:11 |
| jaypipes | efried: so as soon as I verify that isn't the case, I'll push it through. | 21:11 |
| efried | cool beans | 21:11 |
| *** tssurya has joined #openstack-placement | 21:12 | |
| jaypipes | efried: heh, so pretty much as I suspected... https://review.openstack.org/#/c/567150/15/nova/api/openstack/placement/objects/resource_provider.py | 21:12 |
| jaypipes | ^^ the _check_traits_for_alloc_requests() method isn't necessary if the SQL is done correctly in get_trees_matching_all(). | 21:13 |
| efried | ahh | 21:13 |
| efried | jaypipes: Okay, so two choices: 1) You fix it up, and we have to find a third reviewer; 2) Tetsuro fixes it up, with time zone pain. | 21:14 |
| efried | jaypipes: If it was anyone other than Tetsuro, I would shy away from #2, but I have faith that he'll turn it around correctly in one shot. | 21:15 |
| jaypipes | efried: lemme check one thing locally on the following patch. | 21:16 |
| *** tssurya has quit IRC | 21:17 | |
| cdent | dansmith, jaypipes, efried : when you have a minute I could do with some input on the desired interface to this non-http placement tool. Beacuse of the way policy is being handled now, deploying the full wsgi stack (with whatever conf you like) and intercepting it is pretty much the simplest thing. The question becomes: what interface do we want to provide to users? Some options: | 21:25 |
| cdent | a) just the url of the service they do the rest with requests | 21:26 |
| cdent | b) some stubbed out methods that themselves use requests | 21:26 |
| cdent | c) the ReportClient | 21:26 |
| cdent | d) your better idea | 21:26 |
| cdent | "do the rest with requests" is my new bumper sticker, tshirt, laptop decoration | 21:27 |
| efried | cdent: Is it possible for it to look like the handler methods? | 21:27 |
| efried | um, sorry, those guys take raw request/response thingies, don't they? | 21:27 |
| cdent | yes | 21:27 |
| efried | Which isn't what I was going for. | 21:27 |
| cdent | so something needs to mediate a bit | 21:27 |
| efried | I like the idea of being able to call it basically the same way we call the REST APIs. | 21:28 |
| efried | is that "do the rest with requests" ? | 21:28 |
| cdent | pretty much | 21:28 |
| efried | in report client we do it via adapter, which is a pretty thin wrapper around requests. | 21:29 |
| efried | If there were a way to do it via adapter, then client code (like the report client) could run against either one just by changing out the setup. | 21:30 |
| cdent | the reason to not do it with adapter is because we'd have to mock out keystone. not necessarily a bad thing, but maybe ugly | 21:30 |
| dansmith | cdent: personally, I would like something like placement.allocations.update(rp_uuid, body) | 21:30 |
| efried | cdent: Well, right, I was thinking a thing that behaves like adapter, but doesn't have to mock out keystone. | 21:30 |
| cdent | dansmith: your suggestion maps to "b", yes? | 21:31 |
| dansmith | cdent: I don't think so | 21:32 |
| dansmith | cdent: you really mean open a socket and use requests to talk to localhost or something? | 21:32 |
| cdent | oh. how so? | 21:32 |
| cdent | dansmith: the wip is using wsgi-intercept (in the same way the funcitional tests do) so that it can host the entire wsgi stack (for sake of context handling). It doesn't have to do that, we can construct things more directly, but it was an easy option. | 21:34 |
| *** edmondsw has quit IRC | 21:34 | |
| cdent | so it's not a real socket | 21:34 |
| cdent | but it's real http | 21:34 |
| *** edmondsw has joined #openstack-placement | 21:34 | |
| dansmith | okay I've never looked at how the fixture works, so I dunno, | 21:35 |
| efried | Not as much a fan of placement.{noun(s)}.{verb}(identifiers, payload) because we're having to invent a whole set of method specs and map them to placement calls. Do they map 1:1? Maybe at first... How hard will it be to keep them in sync moving forward? | 21:35 |
| efried | If we make it look just like actual REST calls, then (in addition to the drop-in-ability mentioned above) we get all updates for free in the shim. | 21:35 |
| dansmith | but it's hard for me to wrap my brain around using it that directly internally | 21:35 |
| dansmith | efried: yeah, I guess I'd hope that if we have a library, we have a real library that provides a nice-to-use interface and not just "bindings" | 21:36 |
| efried | dansmith: The "placement client" we've been trying to avoid... | 21:36 |
| dansmith | that would be my preference.. something pythonic and not something , uh, webthonic | 21:36 |
| dansmith | efried: no, not really (IMHO) | 21:36 |
| dansmith | efried: it would be more like the abstraction that the http handlers sit on top of | 21:37 |
| cdent | dansmith: are you imagining that this might _actually_ (as opposed to just might) become a thing beyond its value for /migrator ? | 21:37 |
| dansmith | it's cool if everyone just uses requests for the actual api | 21:37 |
| dansmith | cdent: yeah? | 21:37 |
| efried | I mean, the design of the pythonic front end would/could end up looking just like a hypothetical placement client. Only difference is whether the HTTP layer is in front or behind. | 21:37 |
| efried | cdent, dansmith: Oh, yeah, this is one of the things cinder wanted to have for their usage of standalone placement. | 21:38 |
| dansmith | right | 21:38 |
| efried | definitely applicability beyond migrator | 21:38 |
| cdent | right, but that want is _very_ contingent | 21:38 |
| *** edmondsw has quit IRC | 21:38 | |
| dansmith | until we split you can play fast and loose with this a bit, and a wsgi hack to let us do what we need internally is permissable, | 21:39 |
| dansmith | permissible? | 21:39 |
| dansmith | anyway | 21:39 |
| cdent | and I would hope that going down that road (in process use of placement) would result from quite a bit of consideration | 21:39 |
| smcginnis | cdent: I do need to get the discussion going again. | 21:39 |
| smcginnis | Cinder meeting is tomorrow, FWIW. | 21:39 |
| dansmith | if this becomes an external interface, which I kinda think will be useful for a variety of things (i.e. any nova-manage transitions we have to do in the future with placement), it would be nice to have a more programmatic api module, IMHO | 21:39 |
| cdent | smcginnis: not sure there's a huge rush | 21:39 |
| smcginnis | There certainly doesn't seem to be within the team. | 21:40 |
| cdent | dansmith: I think in order to do what you're suggesting in the right way is to do it as you say: a mediating layer underneath the http handlers | 21:40 |
| cdent | which would be cool and a nice layer to have | 21:41 |
| cdent | but probably more effortful than what we want now? | 21:41 |
| dansmith | yeah, which is what I kinda expected: | 21:41 |
| dansmith | the http layers would do the microversion bits, the policy bits, etc and then call this API, which you could call directly if you're nova-manage or something like that | 21:41 |
| dansmith | cdent: like I said, you have leeway while we're locked together I think, but once we split, I think we need to have the more formal thing | 21:41 |
| cdent | yeah, that would require got a lot of change, we're not super clean | 21:42 |
| efried | eek, microversion handling. | 21:42 |
| efried | in anything like a pythonic interface | 21:42 |
| efried | without having actual microversion args | 21:42 |
| dansmith | efried: that's why I think doing it below that layer is better | 21:42 |
| dansmith | so you're not passing version=1.23 to a method | 21:43 |
| cdent | I'm not sure why would we _have_ to be more formal after the split. After the split you can still from placement import direct; with direct.PlacementDirect(Conf): $do some requests | 21:43 |
| dansmith | sure, but what version do you get? | 21:43 |
| dansmith | right now, we know.. it's the same version as the nova code | 21:43 |
| openstackgerrit | Mathieu Gagné proposed openstack/nova-specs master: Multiple Fixed-IPs support in network information https://review.openstack.org/312626 | 21:43 |
| dansmith | but once they're separate packages, you have to be able to say "nova X requires placement Y" in order to be able to import and use it and have the things you expect to be there | 21:44 |
| cdent | that's true without these proposed changes, based on the way we've been writing the report client | 21:44 |
| dansmith | that's because we're calling across to another service, | 21:45 |
| dansmith | which could be on a different machine, running different stuff | 21:45 |
| dansmith | if we import the code directly from the python namespace, we know right now that we're getting the version from the same git hash we're on | 21:45 |
| dansmith | but once you break that, then we're not | 21:45 |
| cdent | I don't really see that as an issue if "do some requests" includes microversions. If the versions not there, it doesn't work | 21:46 |
| efried | you're okay with these pythonic front end methods having a 'microversion' arg? | 21:47 |
| dansmith | efried: I'm not | 21:47 |
| efried | Yeah, that ooks me out too. | 21:47 |
| dansmith | cdent: if we speak http to it, then that's how we avoid a NameError, yes | 21:47 |
| dansmith | I just don't think we should do that :) | 21:47 |
| dansmith | I think speaking http over this fake adapter thing to just /migrator for the purposes of this transition is icky, but probably not the end of the world, | 21:48 |
| cdent | I guess I've not got as complete a view of this librarization as you dansmith . For me it is simply a way to avoid networked http, not http. | 21:48 |
| dansmith | I just don't think that you should expose that from a python library interface | 21:48 |
| dansmith | well, that's certainly one way to approach it, for sure | 21:48 |
| cdent | Basically I don't think we should be writing services that except to use placement as a library, rather we should be writing them so they normally use placement as a service, and _can_ use it as a library is they have to | 21:49 |
| cdent | in which case, efried's adapter idea is useful | 21:49 |
| cdent | s/except/expect/ | 21:50 |
| efried | I agree it's also a bit ookie to have a python lib that speaks http without going over http, but in this case I think it's the lesser evil, and certainly will be the easiest/fastest to implement. | 21:50 |
| dansmith | for the /migrator thing, I've said I think you can do that for expediency | 21:50 |
| * cdent nods | 21:50 | |
| cdent | I think we should probably have a more in depth conversation at some point about the two different library models, because one of them is a lot of rope (not necessarily a bad thing, but usually worth some deeper chat) | 21:51 |
| openstackgerrit | Chris Dent proposed openstack/nova master: WIP: Spike a direct interface to placement https://review.openstack.org/572576 | 22:36 |
| cdent | jaypipes, dansmith, efried, edleafe: wippity wip wip ^ | 22:36 |
| efried | ack | 22:37 |
| mriedem | i can't figure out what's going on with tetsuro's spec here https://review.openstack.org/#/c/545734/ | 22:51 |
| mriedem | it was abandoned and then restored | 22:51 |
| mriedem | same with https://review.openstack.org/#/c/543805/ | 22:52 |
| mriedem | looks like those are both superseded by https://review.openstack.org/#/c/555081/ | 22:52 |
| openstackgerrit | Merged openstack/nova stable/ocata: libvirt: handle DiskNotFound during update_available_resource https://review.openstack.org/571432 | 23:01 |
| openstackgerrit | Merged openstack/nova master: Fix some wrong urls in doc https://review.openstack.org/571147 | 23:01 |
| *** edmondsw has joined #openstack-placement | 23:03 | |
| *** edmondsw has quit IRC | 23:07 | |
| openstackgerrit | Eric Fried proposed openstack/nova-specs master: WIP: Spec: Handling Reshaped Provider Trees https://review.openstack.org/572583 | 23:17 |
| openstackgerrit | Merged openstack/nova master: [placement] Do not import oslo_service for log_options https://review.openstack.org/572231 | 23:21 |
| *** mriedem is now known as mriedem_afk | 23:25 | |
| *** takashin has joined #openstack-placement | 23:31 | |
| openstackgerrit | Chris Dent proposed openstack/nova master: Optional separate database for placement API https://review.openstack.org/362766 | 23:55 |
| openstackgerrit | Chris Dent proposed openstack/nova master: Isolate placement database config https://review.openstack.org/541435 | 23:55 |
| openstackgerrit | Chris Dent proposed openstack/nova master: Ensure that os-traits sync is attempted only at start of process https://review.openstack.org/553857 | 23:55 |
| *** tetsuro has joined #openstack-placement | 23:56 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!