| *** yingjun has joined #openstack-searchlight | 00:00 | |
| *** bpokorny_ has joined #openstack-searchlight | 00:25 | |
| *** bpokorn__ has joined #openstack-searchlight | 00:26 | |
| *** bpokorny_ has quit IRC | 00:26 | |
| *** bpokorny has quit IRC | 00:29 | |
| *** bpokorn__ has quit IRC | 00:42 | |
| *** bpokorny has joined #openstack-searchlight | 00:43 | |
| *** bpokorny_ has joined #openstack-searchlight | 01:01 | |
| *** bpokorny has quit IRC | 01:05 | |
| *** bpokorny_ has quit IRC | 01:06 | |
| *** bpokorny has joined #openstack-searchlight | 01:30 | |
| *** bpokorny has quit IRC | 01:33 | |
| *** bpokorny has joined #openstack-searchlight | 01:33 | |
| *** bpokorny has quit IRC | 05:35 | |
| *** pcaruana has joined #openstack-searchlight | 07:16 | |
| *** pcaruana has quit IRC | 07:23 | |
| *** yingjun has quit IRC | 09:38 | |
| *** yingjun has joined #openstack-searchlight | 11:21 | |
| *** khushbu_ has joined #openstack-searchlight | 12:55 | |
| *** openstackstatus has joined #openstack-searchlight | 13:38 | |
| *** ChanServ sets mode: +v openstackstatus | 13:38 | |
| -openstackstatus- NOTICE: We have recovered one of our cloud providers, but there is a huge backlog of jobs to process. Please have patience until your jobs are processed | 13:40 | |
| *** khushbu_ has quit IRC | 13:55 | |
| *** sigmavirus24_awa is now known as sigmavirus24 | 14:01 | |
| *** yingjun has quit IRC | 14:08 | |
| *** yingjun has joined #openstack-searchlight | 14:09 | |
| *** sjmc7_ has quit IRC | 14:13 | |
| *** yingjun has quit IRC | 14:13 | |
| *** sjmc7 has joined #openstack-searchlight | 14:13 | |
| sjmc7 | TravT: got a sec? | 15:08 |
|---|---|---|
| TravT | sjmc7 hello | 15:20 |
| sjmc7 | i was idly thinking about adding a ‘total’ count to the facet response (which would solve brian’s request in https://blueprints.launchpad.net/searchlight/+spec/count-endpoint) but the /facets endpoint returns a dict {“resource type”: [facets]} which doesn’t lend itself well to that | 15:21 |
| TravT | oh interesting... i had opened a BP on this last night. | 15:22 |
| sjmc7 | on the fence about a breaking API change versus a new endpoint for it | 15:22 |
| TravT | i forgot about this one. | 15:22 |
| sjmc7 | hahaha | 15:22 |
| TravT | https://blueprints.launchpad.net/searchlight/+spec/resource-summary | 15:22 |
| TravT | i'll close that one. | 15:23 |
| sjmc7 | :) | 15:23 |
| TravT | okay, let me think about this for a second... | 15:23 |
| sjmc7 | including it in facets seems neat (maybe allow flags to turn off one or the other) but would mean a breaking change to that API | 15:23 |
| sjmc7 | could also add a new endpoint that could include the facet payload. or just keep them separate | 15:24 |
| TravT | yeah, including in the full facets output also was something i pondered.... | 15:27 |
| TravT | if we really don't want to break that output, there could be a special facet | 15:28 |
| TravT | eg doc_count is just a facet | 15:28 |
| TravT | but that's kinda weird | 15:28 |
| sjmc7 | yeah, that seems nasty | 15:28 |
| TravT | i'm trying out a few requests against API to refresh my memory... just a sec | 15:29 |
| sjmc7 | i’d probably vote against adding it to /plugins because that information’s meant to be more static | 15:31 |
| TravT | i don't know if it is magnum, but my devstack is barely responding to me. sorry, just a sec | 15:31 |
| sjmc7 | your computer has gremlins | 15:31 |
| TravT | so | 15:36 |
| TravT | unless i'm doing something wrong here | 15:36 |
| TravT | it seems that the current facets aren't limited based on all projects or not? | 15:37 |
| sjmc7 | umm… that shouldn’t be the case | 15:37 |
| TravT | let me re-issue token | 15:37 |
| TravT | and try again | 15:38 |
| TravT | i probably horked it | 15:38 |
| TravT | yep, i did | 15:39 |
| TravT | so, if we left in existing facets | 15:39 |
| TravT | 1) break api like you said and maybe go with nested structure | 15:40 |
| TravT | 2) add a special facet for count | 15:41 |
| TravT | 3) add a "summary" as a peer to the resource types | 15:41 |
| sjmc7 | for 1) i’d go with {“rt”: {“facets”: [facets], “count”: 10} | 15:41 |
| sjmc7 | for 3) i’d go with something like you had in the dupe BP | 15:41 |
| sjmc7 | doing it as part of facets seems like it fits best | 15:41 |
| TravT | there is a downside to it | 15:42 |
| TravT | all i have to go on is the current interaction i built into the ui. | 15:42 |
| TravT | but | 15:42 |
| TravT | right now, i can just get plugins | 15:42 |
| TravT | and i don't request facets until a specific one is chosen | 15:42 |
| sjmc7 | ah, interesting | 15:43 |
| TravT | if we were to support showing doc count in say the type dropdown | 15:43 |
| sjmc7 | in my head i imagined that type was just another facet, but that makes sense | 15:43 |
| TravT | eg. type -> instances (42), images (32), etc | 15:43 |
| TravT | this requires doing a full facet hit on every single type proactively | 15:44 |
| sjmc7 | though perhaps if it were part of /facets you’d request /facets?include_fields=false and then /facets?doc_type=OS::Glance::Image | 15:44 |
| *** khushbu has joined #openstack-searchlight | 15:45 | |
| TravT | in some ways this may negate needing to call the "plugins" endpoint at all | 15:45 |
| sjmc7 | yeah, it would in that case | 15:45 |
| sjmc7 | although in my head the parent/child type stuff belongs there | 15:46 |
| sjmc7 | the more static stuff you could reasonably cache for ages | 15:46 |
| TravT | yeah, that's what happens right now | 15:46 |
| sjmc7 | so let’s keep that distinction | 15:46 |
| sjmc7 | question i guess is whether a separate ‘summary’ kind of endpoint might include other useful stuff that wouldn’t make sense under ‘facets' | 15:47 |
| TravT | i also keep facets for a specific resource type cached until you completely remove that resource type from the selection. | 15:47 |
| TravT | if you re-select that type, i hit the facets api again | 15:47 |
| sjmc7 | yeah, that’s fine. the total doc_count is pretty tied to the facet counts | 15:47 |
| TravT | yeah, that's a good question | 15:48 |
| sjmc7 | ideally 2+2 should equal 4 | 15:48 |
| TravT | a summary endpoint might eventually contain many useful queries... | 15:49 |
| TravT | but then if they get too customized, just exposing aggregations directly may make sense | 15:49 |
| TravT | but since we abstract resource types, i'm not too sure about that | 15:49 |
| sjmc7 | yeah, i think there’s an argument for that too | 15:49 |
| sjmc7 | i still think a total count makes sense in /facets | 15:49 |
| sjmc7 | it’s very related | 15:50 |
| *** lakshmiS has joined #openstack-searchlight | 15:51 | |
| sjmc7 | if we’re gonna break the API now would be the time to do it :) | 15:51 |
| TravT | well, we'll have to update the python client and the ui | 15:51 |
| sjmc7 | yeah. a halfway option would be to add a flag to use the ‘new’ behavior | 15:52 |
| sjmc7 | and deprecate it through this release | 15:52 |
| TravT | or we could use microversions | 15:52 |
| * TravT shudders | 15:52 | |
| sjmc7 | :( | 15:52 |
| *** khushbu has quit IRC | 15:53 | |
| sjmc7 | mental note: never return flat lists from APIs :) | 15:53 |
| TravT | i could say a summary flag | 15:53 |
| TravT | yeah, alway objectify | 15:53 |
| TravT | s | 15:53 |
| sjmc7 | yeah, maybe a flag is the best compromise | 15:54 |
| sjmc7 | we can immediately mark it as deprecated | 15:54 |
| TravT | facets?summary=true&fields=false | 15:54 |
| TravT | so, leave as is | 15:54 |
| TravT | but when summary is a flag | 15:54 |
| TravT | add a "summary": {'resource_type': {'count': 29)} | 15:55 |
| TravT | to the top level | 15:55 |
| TravT | or just add a summary to each facet | 15:55 |
| TravT | oops | 15:55 |
| TravT | that would break it to add to each facet | 15:55 |
| sjmc7 | {“rt”: {“count”: 29, “facets”: []}} | 15:55 |
| *** khushbu_ has joined #openstack-searchlight | 15:56 | |
| TravT | {“rt”: {"summary": {“count”: 29}, “facets”: []}} | 15:56 |
| sjmc7 | maybe call it ?totals, but if either is specified use the object response. it’s a little bit nasty smelling but at least it’s backward compat | 15:56 |
| *** bpokorny has joined #openstack-searchlight | 15:57 | |
| TravT | if we were to just break it. now is the time, but I think we'd want the include_fields=false in any event | 15:58 |
| TravT | i *think* the current way I do it above makes sense for a performance standpoint and data currency standpoint | 15:59 |
| TravT | only get the full facets for a type when requested. | 15:59 |
| sjmc7 | yeah | 15:59 |
| sjmc7 | ok. let’s leave the current behavior if no flags are specified just for this release but modify the client and UI to the new behavior; after that we’ll switch to the dictionary response in all cases | 16:00 |
| TravT | i was thinking about horizon overview pane as well and thinking it would be great it searchlight could provide a lot of the summary | 16:00 |
| sjmc7 | yep, that’s where brian was coming from | 16:00 |
| TravT | let's see from the client perspective | 16:00 |
| TravT | openstack search facet list --summary | 16:01 |
| TravT | that doesn't ring quite as nice as openstack search summary | 16:01 |
| sjmc7 | no. but the new default would be to include it | 16:03 |
| sjmc7 | you could specify —no-fields to get the short version | 16:03 |
| TravT | ahh, opposite it | 16:04 |
| sjmc7 | i guess i’m branding it as a facet count for the type rather than a separate piece of information | 16:04 |
| TravT | it does seem like a piece of metadata | 16:06 |
| TravT | that seems reasonable. | 16:06 |
| TravT | there is a "total" piece of metadata on all query requests today | 16:07 |
| sjmc7 | ok. let’s go with that | 16:07 |
| TravT | yeah, that works for me. | 16:07 |
| *** khushbu_ has quit IRC | 16:24 | |
| sjmc7 | no responses about elasticsearch 2 yet | 16:58 |
| TravT | this is a tough week. | 17:01 |
| TravT | people going to the summit are usually running around with their hair on fire | 17:01 |
| sjmc7 | i shall light mine ceremonially on friday | 17:02 |
| TravT | lol. | 17:02 |
| TravT | i'd be happy to do that for you when you arrive, if you'd like | 17:02 |
| sjmc7 | oopa! | 17:02 |
| openstackgerrit | Merged openstack/searchlight: Fix unicode error when booting instance from volume https://review.openstack.org/306217 | 17:15 |
| *** bpokorny has quit IRC | 19:00 | |
| *** sigmavirus24 is now known as sigmavirus24_awa | 19:06 | |
| openstackgerrit | Lakshmi N Sampath proposed openstack/searchlight-specs: WIP: Policy driven RBAC https://review.openstack.org/308010 | 19:34 |
| *** sigmavirus24_awa is now known as sigmavirus24 | 20:06 | |
| *** bpokorny has joined #openstack-searchlight | 20:30 | |
| openstackgerrit | Lakshmi N Sampath proposed openstack/searchlight-specs: WIP: Policy driven RBAC https://review.openstack.org/308010 | 20:58 |
| *** jaimguer has quit IRC | 21:17 | |
| *** lakshmiS has quit IRC | 21:19 | |
| *** sigmavirus24 is now known as sigmavirus24_awa | 22:02 | |
| openstackgerrit | Travis Tripp proposed openstack/searchlight: Add cleanup in functional tests https://review.openstack.org/295156 | 22:06 |
| *** yingjun has joined #openstack-searchlight | 23:05 | |
| *** david-lyle_ has joined #openstack-searchlight | 23:14 | |
| *** david-lyle has quit IRC | 23:14 | |
| *** yingjun has quit IRC | 23:32 | |
| *** yingjun has joined #openstack-searchlight | 23:33 | |
| *** yingjun has quit IRC | 23:37 | |
| *** david-lyle_ is now known as david-lyle | 23:57 | |
| *** lakshmiS has joined #openstack-searchlight | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!