15:03:24 #startmeeting openstack search 15:03:24 Meeting started Thu Jul 23 15:03:24 2015 UTC and is due to finish in 60 minutes. The chair is nikhil_k. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:03:26 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:03:28 #chair TravT 15:03:28 The meeting name has been set to 'openstack_search' 15:03:29 Current chairs: TravT nikhil_k 15:03:34 thanks, nikhil_k 15:03:43 o/ 15:03:46 #topic agenda and roll call 15:03:47 o/ 15:03:48 o/ 15:03:49 o/ 15:03:49 o/ 15:03:54 hey TravT 15:03:58 o/ 15:04:16 agenda here: https://etherpad.openstack.org/p/search-team-meeting-agenda 15:04:22 add anything to it that you'd like 15:05:04 #topic Horizon integration update --> Endpoint in Keystone (include v1 or not?) TravT, sjmc7 15:05:19 so, sjmc7 and i are at the horizon mid-cycle 15:05:28 o/ 15:05:46 my focus has been trying to get images table angularized and getting a widget called magic search on top of it. 15:06:01 this lays the groundwork for utilizing searchlight 15:06:20 and yesterday sjmc7 started porting over our concept patch that we showed at the summit 15:06:37 very early, but we did encounter one question 15:06:43 that is the endpoint in keystone 15:07:01 currently devstack puts in 'host:9393/' 15:07:09 which means the client has to append v1/ 15:07:21 there seems to be about a 50/50 split with how other services do it 15:07:30 in the endpoint list for other services there doesn't seem to be consistency 15:07:53 for example, cinder has a totally different service listed 15:07:58 for their v2 15:08:16 hmm, yeah 15:08:21 so, we were wondering if we should keep the endpoint in the service catalog to not have the 'vX' appended 15:08:30 and force the user of the service to add that 15:08:39 or if we should have the endpoint in the catalog 15:08:48 especially since we dpn't have a client to do that auto-versioning thingy the python clients seem to have 15:08:48 have the vX on it 15:09:22 we should have it unversioned 15:09:34 that's where other projects are trying to move to 15:09:52 seems like cinder is good way to solve it since it also tells clients what versions are supported as endpoints 15:10:12 == kragniz 15:10:24 that's 2 - 1. any other takers? :) 15:11:21 we could put this out on the mailing list and announce our intention to not have it versioned in the endpoint list 15:11:32 guys 15:11:35 hey guys :9 15:11:43 howdy ekarlso! 15:12:21 unversioned with discovery is the typical pattern 15:12:28 no /vFOO in the endpoints :P 15:12:41 at least what we did for designate 15:12:45 allright, we'll stick with the unversioned for now. that's what the current horizon patch assumes. 15:12:45 we have no client 15:12:50 and thus no autodiscovery 15:13:12 sjmc7_: u can easily do that by using the adapter / session stuffi n ksauth 15:13:30 we don't have a discovery api right now (unless I missed something), so do we need to add that? 15:13:54 TravT: u can pass a endpoint too to ks if you want to with a token 15:14:00 typically you put in the requested version and client tries, right? 15:14:27 let's put this on the mailing list and get a decision by next week 15:14:28 TravT: yeah it'll try discovery if you dont pass a explicit endpoint 15:14:51 ok, well, let's stick with no version in keystone then. 15:15:04 so next topic is bug review 15:15:17 #topic Ceilometer / Notifications: https://review.openstack.org/#/c/202392 15:15:28 lakshmiS: any updates here? 15:15:37 after several discussions/reviews we have the patch in decent shape now 15:16:22 haven't seen any further -1's and works fine with ceilometer running 15:16:46 ok. I'm guilty of not doing reviews in about a week, but tomorrow i'm intending to get on that. 15:16:46 yeah, i'm good with it too, thanks for addressing my suggestions 15:16:51 today is last day of horizon meetup 15:17:49 ok, on to next 15:18:01 any other bugs anybody needs discussion? 15:18:08 https://bugs.launchpad.net/searchlight/+bug/1469336 15:18:10 Launchpad bug 1469336 in OpenStack Search (Searchlight) "Errors are returned as html" [Medium,New] - Assigned to Ian Cordasco (icordasc) 15:18:25 we hit this doing the horizon integration and it's a pain 15:18:41 doesn't need discussion i don't think 15:19:15 ok 15:19:31 #topic Blueprint review (Please add any blueprints needing attention below) https://blueprints.launchpad.net/searchlight 15:19:57 ekarlso, there were several the last two weeks you had added but i don't think we discussed 15:20:30 ekarlso, can you re-share the info for setting up designate in devstack 15:20:31 https://blueprints.launchpad.net/searchlight/+spec/designate-plugin 15:20:48 i'd like to get that setup so i can properly review and get that patch in 15:21:53 http://paste.ubuntu.com/11925639/ 15:21:57 ekarlso: I would like to review that but what setup needs to be done? 15:22:06 that's the localrc 15:22:19 does it setup sample data as well? 15:22:32 u just clone searchlight to /opt/stack/searchlight, checkout the PS 15:22:40 ehm 15:22:45 lemme write up instructions :p 15:23:03 well, if there are eisting designate instructions we can follow that 15:23:23 i was talking to a guy at time warner about desginate 15:23:34 he said he's happy it'll be in searchlight 15:23:50 but also asked "will it give me any more data than I get from the CLI" 15:23:54 i'll try and test the designate patch tomorrow or monday as well, be great to get that merged 15:24:01 i didn't know what to tell him? 15:25:06 is there anything else on the designate plugin we should know for reviewing it? 15:25:15 if so, just add a comment on the patch i guess. 15:25:21 i will get to it tomorrow or monday 15:25:43 a big monkey on our back right now are nova instances 15:25:48 #topic https://blueprints.launchpad.net/searchlight/+spec/nova-instances-plugin 15:25:50 o/ 15:26:00 yeah, any reviews on that patch are welcome 15:26:05 nikhil_k have you had any chance to look at that one? 15:26:19 i also intend to look at it tomorrow or Monday 15:26:37 TravT: I did take a look and I think we should be good to go ahead on that based on the initial feedback. 15:26:47 it seems if it is a reasonable start and functions that we could get it in and do additional iteration. 15:26:49 TravT: but I need to test it functionally :) 15:26:58 cool. 15:27:06 woo. yes, please test it. let me know if you have questions 15:27:42 TravT: btw, if you happen visit last week's meeting and channel logs for thursday, I did some sharing on thoughts on what could be potential enhancements needed 15:27:57 https://etherpad.openstack.org/p/designate-searchlight 15:28:04 ^ TravT sjmc7_ lakshmiS 15:28:07 ekarlso: thanks! 15:28:15 great, thanks 15:28:35 gtg, house maintenance on the works! 15:28:35 ekarlso: will ping you on irc if needed. whats your timezone? 15:28:39 nikhil_k, I did read the logs. I guess anything needed to be kept forever could be added to the BP. 15:28:51 uhm, Norwegian / GMT+2 I believe :P 15:28:58 ekarlso. thanks! we'll get your stuff going! 15:29:00 ok good for me 15:29:41 there are two more BP listed on the meeting agenda 15:29:51 BP: https://blueprints.launchpad.net/searchlight/+spec/openstack-client-usage 15:29:58 whose is that? 15:30:02 TravT: sjmc7_ can we schedule a session during the glance midcycle for some discussion around indexing and nova instances? 15:30:14 curious to know what our participation looks like 15:30:20 uh.. sure 15:30:27 sure, i'll be around all next week 15:30:35 cool 15:30:52 melissa can possibly join us 15:31:12 ok. i had a call with her a couple weeks ago. 15:31:32 she seemed very interested in utilizing searchlight 15:31:54 i think you and rosmaita will have the most context 15:32:33 nobody seems to own the client bp? 15:32:48 guess we can move onto the named queries bp 15:32:49 https://blueprints.launchpad.net/searchlight/+spec/named-queries 15:32:55 i am working on that 15:32:55 TravT: we did a intial discussion yday but not so much on instances. I wanted to get some feedback on the use cases to see how we can proceed for long term indexing on that 15:33:22 nikhil_k, she has a very strong interest in utilizing searchlight for cross resource searching as well 15:33:40 TravT: yeah, that part 15:33:46 to help facilitate a concept like amazon resource groups. 15:33:49 + cross region 15:34:06 yeah, we really could spend some time brainstorming on cross region. 15:34:23 turtles all the way down, may be 15:34:44 it would be good to have a working vidyo session on some of these topics 15:35:04 if we could get a few more of these basic patches in. 15:35:52 TravT: so I think my misunderstanding was around the indexing through command line. Does HP use cases never needs direct db indexing? 15:36:20 i don't think a use case ever dictates how you index 15:36:49 but we're not aware of that right now. 15:36:50 TravT: becasue for some API calls like the ones that have potentially large queries -- changes-since can be really expensive indexing operation and in some cases is 413'ed internally 15:37:10 we're not habitually doing a changes-since query though 15:37:17 pagination would take forever to index it, may be days! 15:37:44 sjmc7_: yeah, I meant deleted stuff that can be potentially retrieved using that 15:37:49 nikhil_k, what's the likelihood of us getting notification in nova to carry the full data we need? 15:38:03 because deleted stuff isn't really exposed via the api 15:38:11 is that a BP that we should open on nova? 15:38:16 (spec) 15:38:45 i don't think i understand what you mean about 'deleted stuff' 15:38:47 TravT: I think I would need to discuss the scope of such notifications with you both. we can do that offline 15:39:08 yeah, that would be good. 15:39:09 sjmc7_: deleted instances, images (for now) (that's soft deleted) 15:39:19 maybe we should have a call about nova because i think my understanding is lagging behind yours 15:39:25 deleted instances send a notification 15:39:31 deleted glance images i don't know about 15:39:40 but i thought they did too 15:39:42 yeah, but old ones won't. anyways we can chat later :) 15:40:01 yes, this would be a good working session / vidyo session 15:40:08 let's schedule that 15:40:12 cool 15:40:36 #action nikhil_k schedule nova design session 15:40:46 apparently that isn't how you do actions 15:41:28 thanks nikhil_k, i also feel like I'm missing some important information here. 15:41:45 TravT: hmm, I thought it was the right way for doing # action 15:41:55 https://wiki.debian.org/MeetBot 15:42:05 going back on the topic of named queries. 15:42:14 lakshmiS, i don't know any exact named queried yet 15:42:18 ok, sure TravT 15:42:40 TravT: I only have some basic quries for now. 15:42:45 i was wondering if for glance if rosmaita, nikhil_k, kragniz might know the most common searches needed or requested 15:42:46 currenlty the queries are part of the plugin but once plugin config file patch lands I can move it to the config file 15:43:15 putting in config file will easier to add queries later easily 15:43:54 lakshmiS, there isn't a plugin specific config patch up yet, is there? 15:44:22 1. /images/details?name=foobar 15:44:52 Travt: haven't seen one but heard someone was working on it. ekarlso? or may be no one is working on it 15:44:52 TravT: name, name wildcard, instance_uuid, base_image_ref 15:44:55 2. /images/details?updated_at=when?? 15:45:03 prob wildcards on all those i guess 15:45:18 for updated_at, are time ranges important? 15:45:25 probably 15:45:28 eg between X dates or after X dates 15:45:33 yes 15:45:46 forget the wildcards on UUID, don't know what i was thinking there 15:46:01 rosmaita: :D 15:46:11 probably the operating system 15:46:11 let's stick these on the BP or an etherpad 15:46:12 and version 15:46:59 rosmaita, is that just a well known property? i didn't think that was in the common images properties list 15:47:26 i am pretty sure those are "common image properties" 15:47:37 though i don't know that anyone uses them much 15:47:46 * TravT admits it has been a few months since reading those 15:47:47 but if they are indexed, that will motivate people! 15:47:55 they'll all be indexed! 15:47:59 everything is indexed 15:48:37 http://docs.openstack.org/developer/glance/common-image-properties.html 15:48:53 I think we may need some on the min disk, min ram too 15:49:13 rosmaita: nikhil_k: could you add the notes above to the blueprint? 15:49:16 https://blueprints.launchpad.net/searchlight/+spec/named-queries 15:49:39 ok 15:49:49 one thing i want to see about adding to horizon (although i know you public cloud rackspace guys don't care) is a drop down list of common queries 15:49:57 and we'd pull that from these named queries 15:50:25 thanks guys. 15:50:37 #topic liberty 2 release plan 15:50:41 i should have the named queries patch up by tomorrow 15:50:52 thanks lakshmiS 15:51:04 so, we don't have a release plan yet 15:51:23 * TravT slightly embarrassed 15:51:41 i think the action is for nikhil_k and I to talk offline about that 15:52:00 yes 15:52:21 i'm guessing we get questioned on that later today, because we'll be talking about searchlight for a small bit in the horizon meetup this afternoon 15:52:29 TravT: sjmc7_ and I agreed last week that it would be good to release after nova instances merges 15:52:35 ok. 15:52:58 so, let's make it a goal that we have that and designate plugin merged by next meeting. 15:53:14 assuming they are ready of course 15:53:31 #topic open discussion 15:53:59 I added a section for glance+searchlight in the glance's midcycle etherpad #link https://etherpad.openstack.org/p/liberty-glance-mid-cycle-meetup 15:54:23 #action all interested: please add name and timezone info so that we can try to schedule it accordingly 15:54:37 or if we need to split it 15:54:45 Thanks, nikhil, just added me name 15:55:06 lakshmiS: i put some notes on the "whiteboard" on your BP 15:55:25 rosmaita: saw that. thanks 15:55:59 i have a general discussion question, viz., what are people thinking currently about how to do multi-region searchlight? 15:56:38 we briefly spoke at summit, but haven't come up with a concrete plan. 15:56:45 i think initial idea was having multiple index one for each region 15:56:50 we had informally said at some point that we'd keep a searchlight in each region, and then mabye a super-serachlight that would index them all 15:57:02 yeah, searchlight of searchlights 15:57:10 *shudder* 15:57:16 searchlight++ 15:57:19 authentication is a problem 15:57:31 authenticate is always pain 15:57:46 i *think* tokens are always region scoped 15:57:54 specially across resource types 15:58:06 that's maybe something we should have a proper discussion about, maybe next week 15:58:20 figure out what we want it to look like and then if we can do it 15:58:31 sounds good 15:58:46 agreed... also maybe get a few high level ideas up and get some ML input as well. 15:59:01 anything else for today? 2 mins left 15:59:01 if we're done with that, we've tried to discuss this a couple of times, but if people could give some thought to the implemenation around authenticating the service clients i would welcome suggestions; i'm not sure the current method is ideal 15:59:36 currently SL requires admin creds that it uses to intiailize the various clients, and tries to cache a successful auth when the listener's running 16:00:07 it works and it's not insecure but i'd appreciate someone taking a look at it since it was diret from the summit prototype which i didn't put a huge amount of thought or optimization into 16:00:21 ok, plz link patch / bp 16:00:26 and we'll call it a meeting 16:00:48 Thanks 16:01:22 https://blueprints.launchpad.net/searchlight/+spec/openstack-client-usage 16:01:24 thank everybody. 16:01:28 #endmeeting