15:00:26 <sjmc7> #startmeeting openstack search
15:00:27 <openstack> Meeting started Thu Dec  1 15:00:26 2016 UTC and is due to finish in 60 minutes.  The chair is sjmc7. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:29 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:31 <openstack> The meeting name has been set to 'openstack_search'
15:00:32 <matt-borland> o/
15:00:36 <TravT> o/
15:00:39 <rosmaita> o/
15:00:40 <yingjun_> o/
15:00:41 <lei-zh> o/
15:00:49 <sjmc7> morning
15:01:24 <sjmc7> https://etherpad.openstack.org/p/search-team-meeting-agenda
15:02:00 <sjmc7> probably not going to be a very long one today, will get started
15:02:33 <sjmc7> #topic Testing infra fixed
15:02:55 <sjmc7> I worked with Andreas Jaeger a bit last week to sort out builds for our stable branches
15:03:11 <sjmc7> Because infra was installing elasticsearch, it meant only one version was available for a given test job
15:03:26 <sjmc7> And our liberty & mitaka branches weren’t compatible with ES 2
15:03:51 <sjmc7> Infra added a stage to run a script in tools/test-setup.sh if it exists for any project
15:04:06 <sjmc7> So now each branch can install whatever ES version it wants and we don’t have custom infra jobs for searchlight
15:04:18 <TravT> cool!
15:04:27 <sjmc7> Also makes it easier for devstack potentially
15:04:28 <Kevin_Zheng> o/
15:04:45 <rosmaita> yeah, that's really nice
15:05:28 <sjmc7> Yeah.. i gather there had been some other issues with various projects such that they didn’t want to maintain separate dependencies in the infra job definitions, so there we go
15:05:55 <TravT> nice to have that sorted out.
15:06:05 <sjmc7> Yep. And Andreas did most of the work, which was even nicer
15:06:16 <TravT> :D
15:06:20 <rosmaita> he is a very energetic fellow
15:06:50 <sjmc7> :)
15:07:25 <sjmc7> which brings me to..
15:07:27 <sjmc7> #topic Elasticsearch 5 GA
15:07:51 <sjmc7> The 5.0 GA was a few weeks ago - https://www.elastic.co/blog/elasticsearch-5-0-0-released
15:08:59 <sjmc7> I don’t expect any problems with it, but I’ll try to run Searchlight against it to check. We’ll need to decide what to officially support for the ocata release
15:09:26 <sjmc7> We can support 2.x and 5.x but it’s awkward to run tests against both
15:09:27 <TravT> wow: 25% - 80% improvement to indexing
15:10:02 <rosmaita> i guess that's why they went from ES2 to ES5
15:10:03 <lei-zh> yeah, last week I have a share about searchlight at beijing openstack meetup, someone has asked if we are going to support both es 2.x and 5.x
15:10:36 <sjmc7> I think i accidentally ran SL against one of the 5.0 RCs a few months ago
15:10:56 <sjmc7> But looking at the release notes nothing jumped out that would be a problem like the 1.x -> 2.0 move was
15:11:15 <sjmc7> I think the versioning number change was to bring all their product lines in step
15:11:45 <sjmc7> IIRC, v2 indices aren’t compatible with v5
15:11:57 <sjmc7> I think due to changes in Lucene
15:12:11 <rosmaita> i thnk they say 5 can read 2
15:12:16 <rosmaita> but def not 1.x
15:12:44 <sjmc7> OK, I’ll check. There’s a migration tool again for this version jump
15:13:06 <sjmc7> There’s also a new scripting language
15:13:19 <TravT> ipv6 support...
15:13:56 <sjmc7> Yeah, ipv6 is a big one because we can make IP range lookups work properly, so that might be a reason to make a change. I’m hesitant to make the code aware of the ES version though
15:14:18 <sjmc7> There’s also a better/safer scripting language
15:14:20 <TravT> the new data structures sound good.  however, they look like things that we can only use if we were to only support es5... or if we detect es version at indexing time.
15:14:42 <sjmc7> Right - and we could do that reasonably easily
15:15:31 <sjmc7> They also brought back dots in field names :)
15:15:51 <TravT> the improvements in es 5 are quite substantial it seems.
15:16:29 <sjmc7> Yeah, it’s been a  while in the making. I’ll give it a try and see how painful it is; if we can take advantage of some features by detecting version i think i’m ok doing it
15:16:37 <sjmc7> IPv6 support would be nice
15:17:35 <sjmc7> So lei-zh , if anyone asks again tell them probably this release :)
15:18:05 <sjmc7> Moving on..
15:18:07 <sjmc7> #topic Searchlight in os-ansible
15:18:21 <sjmc7> TravT pointed this out yesterday
15:18:34 <sjmc7> Rackspace added a review to add Searchlight to openstack-ansible - https://review.openstack.org/#/c/404419/
15:18:59 <sjmc7> I’d spoken to Neil a bunch of times about it in past summits but it looks like he finally got asked to work on it, which is good
15:20:23 <TravT> yeah, nice to see it happening now.
15:20:26 <sjmc7> Since it increases our audience, if they come looking for help with it in the RC room please assist them
15:21:01 * TravT make note to fix irc connectivity problems plauging him in the office.
15:22:27 <sjmc7> Does anyone here use openstack-ansible?
15:22:55 <TravT> rosmaita: ?
15:23:07 <rosmaita> not me personally, but i know some people
15:23:34 <Kevin_Zheng> no
15:23:36 <sjmc7> We used ansible, but not the openstack project for some reason
15:23:47 <sjmc7> So I may lend a hand
15:24:15 <sjmc7> OK…
15:24:19 <sjmc7> #topic Pipeline patch
15:24:42 <sjmc7> lei-zh uploaded a new version of the publishing/pipeline patch yesterday i think? https://review.openstack.org/#/c/359972/
15:25:02 <sjmc7> There’s been some email discussion about the Zaqar use case with rob from horizon
15:25:19 <robcresswell> \o/
15:25:26 <sjmc7> Ah, hello
15:25:42 <robcresswell> Yeah, I usually lurk in these meetings if I'm around :)
15:26:05 <sjmc7> I’ve been reading if not participating - how’s that going?
15:26:46 <robcresswell> Interns are researching a few approaches, because we're not entirely certain on the data source yet (so they're looking at Panko too)
15:26:52 <robcresswell> and there's a POC by the zaqar PTL
15:27:07 <robcresswell> for doing things like table updates via websocket iirc
15:27:24 <robcresswell> I dont know where you guys are on the publishing side of things?
15:27:45 <sjmc7> We’re trying to get that patch in before o-2
15:28:06 <robcresswell> Ah grand
15:28:31 <sjmc7> It actually came up in another mailing list discussion about sanitizing oslo.messaging data; Zane suggested that functionality might be something other teams are also interested in, which is something we’ve tried to push a bit at summits past
15:29:08 <lei-zh> The pipeline patch is already there, I'm going to add  test cases for it
15:29:11 <sjmc7> It’d be nice if that was more of a central project one day, but that might still be a way off
15:29:27 <sjmc7> I’m going to try to review it today lei-zh
15:29:59 <lei-zh> sjmc7: that's good, early feedbacks are appreciated
15:30:09 <sjmc7> Incidentally, the ocata-2 milestone is Dec 12th - 16th
15:30:11 <robcresswell> It'd be good if something could land; developing on top of multiple WIP patches across multiple services is a bit of a nightmare
15:30:28 <robcresswell> sjmc7: Its usually the Thursday specifically, so 2 weeks today :)
15:30:38 <TravT> yeah, i'd rather see it get it in this milestone.
15:30:38 <sjmc7> hurray
15:31:02 <TravT> otherwise it is too late for folks like rob to make any use of this release
15:31:20 <sjmc7> Agreed; it’s a big priority for us this release
15:31:46 <lei-zh> yeah, we try to settle down the pipeline as soon as possible, after that can focus on the detail of zaqar publisher
15:31:53 <robcresswell> That's good to hear
15:32:25 <robcresswell> the two use cases I outlined in the email thread are updating tables via push and being able to provide desktop style notifications, so live feedback in Horizon.
15:33:04 <robcresswell> (just for anyone who didnt see the thread)
15:33:14 <sjmc7> Yeah, that’d be nice. Zaqar already supports multitenant websockets so if we can feed it appropriately it makes sense
15:33:52 <robcresswell> Yeah, we just need some of the code to settle so we can build on top of it. Even if its basic, since we're only really looking for POC work for now.
15:34:16 <sjmc7> Right, so we won’t bikeshed too much of it since things’ll likely shift around a bit
15:34:32 <sjmc7> Much as i love a good bikeshedding session
15:34:50 <sjmc7> That was all i had on the agenda
15:34:55 <sjmc7> #topic Open discussion
15:34:59 <robcresswell> Yeah, I wouldn't get too hung up over long term issues, at least from our perspective. We're not looking for it to be feature complete first time round :)
15:35:14 <sjmc7> It’s Openstack, nobody expects that :)
15:35:19 <robcresswell> heh
15:35:45 <robcresswell> I have a question
15:35:51 <sjmc7> You may ask your question
15:36:09 <robcresswell> Searchlight-UI; whats the status there? Anything that can/should be pushed to Horizon?
15:36:18 <sjmc7> matt-borland / TravT ?
15:36:45 <robcresswell> I'm still rather keen on making sure we stay on top of that for future work, despite the roadblocks people keep putting down.
15:37:03 <TravT> i think that the flavor stuff in searchlight ui needs to be deleted because we merged that on master.
15:37:04 <matt-borland> robcresswell, I would expect that the resource registrations should move into Horizon.
15:37:15 <matt-borland> and what TravT said :)
15:37:24 <TravT> it would be nice if instances were moved to horizon.
15:37:40 <sjmc7> Yeah, i recall merging some stuff into our tree end of last cycle that maybe should’ve been in horizon
15:38:01 <TravT> the key thing is we don't want to just delete it out of searchlight ui unless it is moved to a default activated panel on horizon.
15:38:10 <robcresswell> TravT: Understood
15:38:16 <matt-borland> yep, and that was why it went into SL-UI
15:38:53 <robcresswell> And then I guess the followup question; what's everyone workload right now? I'm just finishing up create-net and tyr's form-preview patch.
15:39:09 <robcresswell> I could probably find time to push on instances, and I know others are doing Identity/Users.
15:39:15 <TravT> however, with the work we did last release to sort of co-locate new and old work on panels and just switch angular on or not, i think that things can merge on horizon even if angular isn't active because the javascript (and registrations) will still be available.
15:39:53 <TravT> so searchlight ui should still be able to use them even though the panel hasn't been fully turned on...
15:40:34 <matt-borland> Double-registrations may get a bit hairy
15:40:58 <TravT> right, so it is best to delete the registration in searchlight ui and move to horizon.
15:41:02 <matt-borland> I think it's only a problem if we double-register table columns.
15:41:24 <matt-borland> which we could use .replace for instead of .append if we can't think of anything else. :)
15:43:28 <robcresswell> TravT, matt-borland: Do you guys have any time to look at porting some of that content, or should I take that over? I'll probably be mostly free in a week or so. Just before O-2.
15:43:49 <matt-borland> I don't have time to port it over
15:44:02 <robcresswell> Cool, fair enough
15:44:20 <matt-borland> Mostly, the paths have to change, but it's just a lot of code and a few assumptions. :)
15:44:44 <robcresswell> Sure
15:44:54 <robcresswell> I'll take a look then. Thats all from me :)
15:45:02 <matt-borland> thanks robcresswell!!! :)
15:45:04 <sjmc7> Ok, thanks
15:45:07 <sjmc7> Anything else?
15:45:21 <lei-zh> happen to see this part in the doc, http://docs.openstack.org/developer/searchlight/dev-environment.html#the-v2-elasticsearch-client
15:45:30 <TravT> robcresswell: add me to any reviews though
15:45:38 <robcresswell> TravT: Will do
15:45:47 <sjmc7> lei-zh: yeah, we can maybe remove that now
15:46:36 <lei-zh> ok, I'll take a look at it
15:47:02 <sjmc7> Cool, thanks. Ok, if there’s nothing else i’ll end this in a couple of minutes
15:47:23 <robcresswell> Thanks sjmc7
15:48:13 <sjmc7> Ok, thanks folks
15:48:56 <sjmc7> #endmeeting