20:00:19 <robcresswell> #startmeeting horizon
20:00:20 <openstack> Meeting started Wed Jun 22 20:00:19 2016 UTC and is due to finish in 60 minutes.  The chair is robcresswell. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:00:21 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
20:00:24 <openstack> The meeting name has been set to 'horizon'
20:00:32 <david-lyle> o/
20:00:34 <r1chardj0n3s> o/
20:00:36 <brad_pokorny> o/
20:00:37 <ezpz> o/
20:01:07 <robcresswell> Hi all
20:01:18 <tsufiev> o/
20:01:34 <rdopiera> hi
20:01:43 <rdopiera> (I'm back!)
20:01:58 <tsufiev> wow )
20:02:02 <r1chardj0n3s> ohai rdopiera!
20:02:06 <tsufiev> rdopiera, back in Horizon?
20:02:07 <robcresswell> Meeting agenda for today is clear, it seems, so we can just move to an open discussion right away if anyone wants to raise anything to the wider group
20:02:17 <david-lyle> O_o
20:02:48 <robcresswell> #topic Open Discussion
20:03:02 <robcresswell> rdopiera: As in, back on Horizon?
20:03:10 <rdopiera> robcresswell: yeah
20:03:17 <robcresswell> \o/
20:03:26 <tsufiev> great news!
20:03:34 <rdopiera> well, officially in a few weeks, but I'm going to start catching up to what changed already
20:03:52 <tsufiev> rdopiera, in 2 words: A LOT :)
20:03:59 <rdopiera> yeah, I can see
20:04:03 <robcresswell> thats great news
20:04:09 <rdopiera> thanks!
20:04:50 <robcresswell> One thing I'd like to point at is https://review.openstack.org/#/c/332745/
20:05:17 <robcresswell> Its not up for review yet, but if anyone is curious about how schema form works and what a workflow might look like, this is a good one to look at
20:05:31 <david-lyle> you -2'd your own patch, impressive indeed :P
20:05:42 <r1chardj0n3s> that's leadership
20:05:51 <robcresswell> Create net has a lot of settings interaction and intertwined fields, so its a good example of a complex workflow
20:06:03 <robcresswell> Ha, -2 stays between patches, workflow doesnt.
20:07:05 <r1chardj0n3s> robcresswell: the commit message says "The Create Net workflow, for example, goes from 12+ files to 2" yet I don't see 10 files removed...
20:07:23 <tsufiev> r1chardj0n3s, lol
20:07:26 <robcresswell> https://review.openstack.org/#/c/264299/ << its in rference to this
20:07:37 <robcresswell> Which was my previous work that I gave up on.
20:07:59 <r1chardj0n3s> ok
20:08:08 <r1chardj0n3s> (apologies, still waking up)
20:08:19 <robcresswell> No I wasn't clear in the commit.
20:08:44 <robcresswell> The main thing is it allows us to be consistent on all the fields and field validation messages
20:08:45 <tsufiev> robcresswell, I'm going to look into it next week (once I finish debugging horizon performance idiosyncrasic issues)
20:09:05 <robcresswell> Because at the moment, the angular forms are a bit of a duplication nightmare.
20:09:37 <r1chardj0n3s> yep
20:10:19 <robcresswell> I'll talk through it more at the midcycle, but just wanted to get it on people's radar.
20:11:57 <robcresswell> The other thing I wanted ot bring up was JS docs
20:12:08 <robcresswell> But it appears we have none of the HP folk in the meeting.
20:12:50 <r1chardj0n3s> Oh, thanks for the reminder. I need to hassle the guy who volunteered to do that generation.
20:13:18 <robcresswell> Probably thought he'd escaped :p
20:13:27 <r1chardj0n3s> yep!
20:13:52 <tsufiev> a short update on integration tests failure rates: I believe that https://review.openstack.org/#/c/330203/2 should fix 85% of intermittent failures, the plan is to see if it succeeds 6 times in a row - and then consider it working
20:14:25 <robcresswell> Thanks tsufiev
20:14:43 <robcresswell> Was anyone able to look further into npm-run-test failures last week/
20:14:54 <robcresswell> I've been away, so not sure where thats up to
20:16:10 * robcresswell assumes silence means no :)
20:16:12 <david-lyle> robcresswell, I think not much
20:16:16 <tsufiev> I don't recall any progress regarding npm-run-test
20:16:42 <david-lyle> krotscheck, provided a link on instruction on how to better debug
20:16:50 <david-lyle> but he's been pulled to other things
20:17:06 <david-lyle> https://krotscheck.net/2016/06/01/how-to-simulate-an-openstack-infra-slave.html
20:17:23 <david-lyle> but volunteers were in short supply
20:17:25 <robcresswell> I'll try and rearrange things next week to get some time to look at that. The test failures are a pain.
20:17:31 <tsufiev> btw, did anyone look towards his krotscheck new initiative - js-openstack-lib?
20:17:41 <r1chardj0n3s> wasn't aware of it
20:17:44 <robcresswell> I'm away Thurs/Fri this week, but then I'm all done with holidaying.
20:18:21 <robcresswell> david-lyle: Thanks btw
20:18:22 <tsufiev> they have regular meetings on Wednesdays, see http://eavesdrop.openstack.org/#JavaScript_SDK_&_Tooling_Meeting
20:19:31 <tsufiev> I'm wondering how much attention should we pay to it, to understand if it evolves in a direction... that could be reused in Horizon in some future
20:19:40 <tsufiev> say, when we give up on Django :)
20:19:48 <r1chardj0n3s> it seems there's no Horizon folks in that at all
20:20:30 <r1chardj0n3s> what's the mandate of the group?
20:20:55 <tsufiev> JS SDK for everyone, to rule them all
20:21:06 <robcresswell> Oh, it looks like the meetings only started last week
20:21:27 <r1chardj0n3s> to mirror the Python openstacksdk?
20:21:35 <tsufiev> I think so
20:21:39 <r1chardj0n3s> ok
20:22:26 <tsufiev> I still have doubts if Django API wrappers can handle a sufficient number of concurrent browser connections, each having Angular JS runtime firing dozens of requests simultaneously
20:22:45 <tsufiev> and if it can't, js-openstack-lib could be the thing that we need
20:23:22 <r1chardj0n3s> tsufiev: but a browser won't fire "dozens of requests simultaneously" ... all current browsers throttle to around half a dozen
20:23:30 <robcresswell> I think if we get to a stage where the API layer is the constraining factor, then I'd be quite happy actually :)
20:23:45 <tsufiev> r1chardj0n3s, then imagine there are 50 browsers concurrently asking for some stuff
20:24:01 <r1chardj0n3s> tsufiev: yep
20:24:27 <r1chardj0n3s> I remain dubious about CORS being something operators want to have to deal with
20:24:28 <tsufiev> robcresswell, these moments seem to happen very suddenly :)
20:24:45 <r1chardj0n3s> but we'll see how it plays out I guess
20:24:58 <robcresswell> This is true. I just think talking about replacing the API layer is rather premature optimisation in this case.
20:25:12 <tsufiev> robcresswell, like our current setup with Horizon + Contrail behind Neutron, where to render project/networks page with a small number of resources it takes 26 seconds for Horizon
20:25:27 <tsufiev> becuase it seems to fetch half of the underlying cloud resources
20:25:51 <robcresswell> Yep
20:25:55 <tsufiev> by our I meant some downstream installation
20:26:25 <robcresswell> Just comes back to reviewing and being wary of any api calls being made, and paying attention to what can sensibly be cached
20:26:44 <robcresswell> If you have suggestions, we can discuss at the midcycle too
20:27:34 <robcresswell> tsufiev: I imagine your performance investigations will throw up some interesting optimisations for all of horizon too :)
20:28:21 <robcresswell> Perhaps we could repurpose https://github.com/openstack/horizon/blob/master/doc/source/faq.rst for notes on those config issues too.
20:28:25 <tsufiev> robcresswell, I'm crossing fingers on my every hand and leg the result of these investigations will not be 'do not use horizon' :D
20:30:22 <robcresswell> david-lyle, r1chardj0n3s: At the summit, OSIC were talking about running tests on large-ish deployments for performance data
20:30:41 <robcresswell> Could we leverage that at all, and nab some logs on Horizon?
20:30:49 <tsufiev> robcresswell, I heard this thing in Tokyo :P
20:31:15 <r1chardj0n3s> I made enquiries about that back then but heard nothing useful, with no follow-up I'm afraid
20:31:19 <david-lyle> we need a formal request to use the cluster
20:31:34 <r1chardj0n3s> my formal request to use the cluster is still in the queue it seems
20:31:58 <david-lyle> oh
20:32:06 <david-lyle> well then I see great success
20:32:15 <tsufiev> by the way, the original plan for performance testing was to combine osprofiler middleware and rally benchmarks using fake resources
20:32:39 <david-lyle> fake resources would miss most of the point IMO
20:32:46 <david-lyle> it's the API calls that kill us
20:32:46 <tsufiev> we're still in the middle of migrating osprofiler to new mongodb driver, which should speed up data retrieval on horizon side
20:33:33 <tsufiev> david-lyle, yes, it would be closer to 'performance sanity testing' - speaking of a reasonable number of calls
20:33:48 <tsufiev> still something to start with :)
20:34:14 <david-lyle> better than now, for sure
20:34:43 <tsufiev> anyways, I hope that we'll make some progress in osprofiler/horizon area by the beginning of midcycle
20:35:24 <robcresswell> Sounds good
20:35:46 <robcresswell> Any other efforts that need discussion/ have been bottlenecked by reviews etc?
20:36:59 <robcresswell> Cool, lets call it there. Thanks everyone!
20:37:10 <robcresswell> #endmeeting