15:00:53 #startmeeting openstack search 15:00:53 Meeting started Thu Mar 16 15:00:53 2017 UTC and is due to finish in 60 minutes. The chair is sjmc7. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:54 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:56 The meeting name has been set to 'openstack_search' 15:01:31 o/ 15:01:33 Morning all 15:01:37 o/ 15:02:07 * gibi is lurking 15:02:26 Kevin_Zheng_Hex, your new irc nic looks cool 15:02:56 Hah, I cannot use irccloud at home, so I use HexChat 15:03:13 :) 15:03:24 Travis is out this week 15:03:38 so we’ll get started. since gibi is lurking: 15:03:39 #topic Nova 15:04:10 i took a look at https://docs.openstack.org/developer/nova/notifications.html and it looks like versioned notifications shouldn’t be too hard to use; the payload is quite similar to the unversioned ones at present 15:04:32 one thing i wondered is whether we can JUST support versioned notifications or we have to support both 15:05:04 the main problem with leaving both paths is that the old one won’t get tested 15:05:14 versioned still missed alot 15:05:34 oh? maybe i didn’t look hard enough :) 15:06:05 the payloads looked quite exhaustive 15:06:07 sorry, you mean for the nova list case or the whole thing? 15:06:09 yeah versioned is not in feature parity with the legacy yet 15:06:22 i mean in terms of searchlight receiving them 15:06:26 instance is almost done 15:06:27 but if you have a list what is the most important thing missing then we can push that 15:06:36 but overall still missed alot 15:07:04 ok. if there are specific fields that are missing (esp from instances) then file a bug/bp 15:07:13 BTW, I noticed that instance.update didnt send out versioned notification 15:07:30 and the legacy one didnt include description 15:07:45 gibi: what’s your feeling on how people deploy nova? are they switching to versioned notifications? 15:07:54 Kevin_Zheng_Hex: we transformed the instnace.update already so you might hit a bug 15:07:59 if there are differences between versioned and unversioned that makes me want to just choose one 15:08:38 sjmc7: by default nova emits both legacy and versioned but I think nobody really switched to consume the versioned one yet 15:08:47 sjmc7: mostly due to the missing feature parity 15:09:08 yeah, too much to notify :) 15:09:26 you can see the current status of the versioned transformation here https://vntburndown-gibi.rhcloud.com/index.html 15:09:39 this not just includes what is ready but also what has some patch up 15:10:24 ok. for now then maybe we don’t have to switch to versioned very urgently? 15:10:29 anyhow if you can come up a list of important notifications to transform I can make that a priority on the nova side 15:11:08 sjmc7: nova - searchlight integration depends on that searchlight switches to versioned notification 15:11:52 ah, ok. so that means we would need to support both paths? 15:12:19 or force people to enable both, i guess? 15:12:32 sjmc7: by default nova emits both 15:13:52 sjmc7: #link https://github.com/openstack/nova/blob/master/nova/conf/notifications.py#L86 15:14:02 ok.. i’m just trying to figure out whether we can drop support for unversioned or not. it looks like not, given that some types aren’t implemented 15:14:43 sjmc7: please list those types and then I will try to progress with those. But yes if someting is missing today then you cannot just use the versioned today but you have to use both 15:14:47 i guess we can begin on making the change and maybe drop the old code base if it’s appropriate later in the release 15:14:48 yeah, at least for a few releases 15:15:17 the nova notification handler is already the most complex code we have, and i fear having two slightly different copies of it is going to be even more complicated 15:15:53 at the very least we’ll begin work to switch over to consuming them, and see how things are going in a few weeks. if there are events that we need missing i’ll let you know 15:16:02 Hmm, actually we have only instance/sg/flavor 15:16:09 handler 15:16:18 yep. and many of those are implemented 15:16:25 flavor is already versioned 15:16:31 instance is almost done 15:16:50 maybe we can only support versioned 15:17:01 sjmc7: cool, thanks! 15:17:35 I mean instance versioned notifications are almost done in Nova side 15:18:26 yeah, we should look at the list carefully and let gibi know if anything’s missing we really need 15:18:34 ok, moving on 15:18:40 #topic BPs tagged for pike 15:19:07 i’ve updated https://blueprints.launchpad.net/searchlight a bit 15:20:05 most of what’s on there is work that’s been discussed, but if anyone’s got other stuff then feel free to tag it for pike or let me know 15:20:26 i’m in the process of adding an ES 5 test job since it’s easy to break things 15:21:53 #topic reviews 15:22:21 https://review.openstack.org/#/c/444095/ is a patch on horizon that is part 1 of removing some code from searchlight-ui 15:22:58 so if anyone’s got time and is able to help test it that would be good. there was a mailing list thread about it last week 15:23:15 I've asked Richard to focus on getting that done for Pike 15:23:54 So its pretty much top of our list really 15:24:10 Ideally we can do it in a way that suits both codebases :) 15:24:28 ok. my time is pretty limited and my expertise lacking but i’ll help as much as i can :) i’ll try to get travis to help out with that one 15:24:36 Added to my list 15:24:40 question, does it mean that horizon plans to use searchlight-ui code to replace nova server pannel 15:24:55 yes 15:25:26 ok, that's cool 15:25:41 even aside from that, that code should never really have been in sl-ui to start with 15:25:47 there'll be some differences though, since we can't force searchlight on people yet. So it'll have to fall back to plain old server filtering. 15:25:48 it was a function of timing and speed 15:25:59 sjmc7: Yeah, we weren't getting things merged fast enough :( 15:26:25 there wer emore important things at the time. it’s good it’s happening now 15:26:28 Long term, I'd like to get to a stage where we can just tell people to install searchlight though 15:26:45 if nova gets to a point where its experience is much better with SL that’ll be easier 15:27:19 yeah 15:27:20 i’d really like to see the kind of thing we demoed in vancouver sometime, the context-less nav bar search 15:28:59 ok, so any help with that much appreciated. and thanks robcresswell for prioritising it 15:29:43 I'm a bit less sold on the search-first idea, but yeah, definitely like the faceted search and speed of it. 15:29:54 and you're welcome. Its long overdue on my part. 15:30:02 yeah, me too :) but i like the idea of being able to find stuff quickly 15:30:10 the other review i wanted to mention was https://review.openstack.org/#/c/444068/ which Kevin_Zheng_Hex put up to make searchlight client a bit more user-friendly 15:30:57 essentially extracting [‘hits’][‘_source’] from the ES results 15:31:12 i’ve no objection to doing it if it helps consumption 15:31:42 in many cases the summary and metadata attached to each result is not useful 15:31:50 ok, that was all i had on the agenda 15:31:55 #topic open discussion 15:32:11 If anyone’s got anything else, you may have the floor, otherwise i’ll call it in a few minutes 15:32:30 Is there any other content we should be looking at porting to Horizon? 15:32:51 full panels/widgets etc 15:35:08 I think that is a no :) 15:35:54 robcresswell: wanna know some details about how horizon consuming data from zaqar 15:36:27 robcresswell: sorry, connection dropped. i think there was a a lot of work done for glance in SL-ui so maybe worth a look 15:37:11 sjmc7: Interesting, we're actually due to remove the python glance implementation entirely in Horizon, since angular replaced it 2 cycles back I *think* 15:38:05 splendid! 15:38:18 I can take a look and see if there is content we should be using though 15:38:46 lei-zh1: Best bet is to email the guys from the earlier thread. I can't remember the addresses from the top of my head - Calvin and Laurent 15:38:48 i may be misinformed 15:40:29 robcresswell: yeah, I'll email them to get more detail, just forget their irc name : ) 15:44:07 ok, if there’s nothing else, let’s call it here 15:44:27 have a good morning/afternoon/evening/night 15:44:52 thanks, bye 15:45:35 #endmeeting