08:00:36 <robcresswell> #startmeeting horizon
08:00:36 <openstack> Meeting started Wed Jun  1 08:00:36 2016 UTC and is due to finish in 60 minutes.  The chair is robcresswell. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:00:37 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
08:00:40 <openstack> The meeting name has been set to 'horizon'
08:01:06 <robcresswell> Morning. Anyone around for Horizon meeting?
08:01:18 <r1chardj0n3s> hullo
08:01:24 * tsufiev waves his hand
08:01:24 <r1chardj0n3s> if you really want to :-)
08:01:33 <ayah> hi
08:01:57 <r1chardj0n3s> I'm also cooking my dinner right at the moment ;-)
08:02:29 <robcresswell> There's no agenda, so just a quick couple of announcements then we can have open discussion
08:03:07 <robcresswell> I'll be doing both the first stable/mitaka maint release, and the Newton Milestone 1 release in tbhe next day or two.
08:03:35 <robcresswell> #link https://launchpad.net/horizon/+milestone/newton-1
08:04:19 <r1chardj0n3s> time flies
08:04:20 <robcresswell> We're a little down on blueprints from where I'd like us to be, but have a ton of bugfixes in
08:04:51 <robcresswell> So thats really good to see. Remember to check the N-2 milestone as I carry things over to make sure we dont miss deadlines.
08:05:40 <robcresswell> In terms of stable improvements, we've already backported more than 2/3 the number of fixes that went into stable/liberty, so thats a great improvement too.
08:06:13 <robcresswell> Finally, another reminder about the midcycle:
08:06:16 <robcresswell> #link https://wiki.openstack.org/wiki/Sprints/HorizonNewtonSprint
08:06:32 <robcresswell> Feel free to add topics to the etherpad.
08:06:57 <robcresswell> #topic Open Discussion
08:07:26 <robcresswell> Oh, and a big thankyou to tsufiev for fixing the gate and chasing the various patches through!
08:07:34 * tsufiev bows
08:07:37 <r1chardj0n3s> oh yes!!
08:07:54 <r1chardj0n3s> nice that we can get patches through again :-)
08:08:10 <tsufiev> I�d like to discuss another potential DOA breaker though ;)
08:08:14 <r1chardj0n3s> ouch
08:08:25 <tsufiev> AKA AVAILABLE_REGIONS renaming thing
08:09:06 <tsufiev> is the overall consensus that wiping out all REGION mentions where they are incorrect is too much for a setting name refactoring?
08:09:38 <tsufiev> here is a link https://review.openstack.org/#/c/224756/
08:09:47 <r1chardj0n3s> I have no opinion on this, sorry
08:10:25 <tsufiev> because if there is an opinion that it is too much, I would reduce the patches impact and make them to only alter just one setting name
08:11:00 <tsufiev> although it may cause some minor confusion in future when endpoints will be referenced as regions
08:11:09 <tsufiev> (referenced in code of DOA)
08:11:19 <robcresswell> I quite like the consistency in that patch, even though its a far reaching change
08:11:34 <robcresswell> Will need a release note though.
08:11:56 <robcresswell> Ah, the horizon patch has one. Ignore me.
08:12:09 * tsufiev stops typing link to horizon patch
08:12:41 <tsufiev> the important thing about these 2 patches is that it is preferrable to land them (if we really aim to) early in the cycle
08:13:20 <tsufiev> that was my message :)
08:13:31 <robcresswell> Okay. I'll take a look this week
08:13:49 <robcresswell> Will ask bpokorny / david-lyle / lhcheng to look too.
08:13:56 <robcresswell> THey tend to have opinions on those things
08:13:58 <tsufiev> I�ll update DOA patch to make tests pass again
08:14:10 <tsufiev> seems there was some badly resolved merge conflict
08:15:09 <robcresswell> tsufiev: Does the horizon change for the integration tests need backporting?
08:15:17 <robcresswell> Strangely they dont seem to be failing on the stable branches
08:15:41 <tsufiev> robcresswell, maybe devstack wasn�t updated for stable branches?
08:15:57 <tsufiev> so the bug is still there, it hasn�t just triggered yet
08:16:30 <robcresswell> Ah, interesting thought
08:16:42 <tsufiev> will doa==2.3.0 be consumed for stable/mitaka?
08:16:44 <robcresswell> Both stable/liberty and /mitaka use uncapped d_o_a iirc
08:17:00 <tsufiev> ah, ok
08:17:34 <robcresswell> Doesnt cause a problem unless someone backports that keystone change. I wonder if we should proactively fix anyway.
08:17:35 <tsufiev> if both liberty and mitaka switched to the latest doa, then it would make sense to backport horizon patch as well, at least for consistency
08:17:41 <robcresswell> Yeah
08:18:17 <robcresswell> tsufiev: Do you have time to backport, or do you want me to?
08:18:25 <tsufiev> robcresswell, will do it
08:18:30 <robcresswell> Thankyou
08:19:52 <robcresswell> Any other questions/ issues? Otherwise we will end early :)
08:20:12 <tsufiev> I had some concern / food for thought / suspicion recently...
08:20:24 <tsufiev> mostly it�s about Django API wrappers and performance
08:20:47 <tsufiev> so we�re switching to many asynchronous client-side calls done via Angular
08:21:09 <tsufiev> but are we sure that Django under Apache can safely handle all these calls?
08:21:33 <tsufiev> I have a feeling that Django is going to become a performance bottleneck in case of many concurrent users
08:22:13 <robcresswell> Hmm.
08:22:14 <tsufiev> and is a concurrent users stress testing a thing that somebody tried ever?
08:23:02 <robcresswell> If it did bottleneck, the worst case scenario would be that it dropped to a synchronous pace, unless it fell over entirely.
08:23:39 <tsufiev> as I read about greenlets and its comparison how Django with Apache limited threads, I got this feeling - look, it is exactly our case!
08:23:40 <robcresswell> As for scale, I don't have any solid data. I know eric is running the new swift UI, and hasnt complained about that particular issue yet
08:24:18 <tsufiev> robcresswell, okay, maybe Horizon has yet to hit these issues :)
08:24:38 <tsufiev> let us not run ahead of time
08:24:54 <tsufiev> r1chardj0n3s, what do you think of ^^ ?
08:25:02 <robcresswell> We may have lost richard to dinner
08:25:27 <tsufiev> better lose dinner to richard :)
08:26:32 <robcresswell> haha
08:26:33 <tsufiev> robcresswell, treat the above as a piece of �food for thought�
08:26:44 <tsufiev> better to start thinking about it early
08:26:50 <robcresswell> Agreed, lets pick it up again when we have more content to test and verify performance against
08:27:20 <robcresswell> Instances table/panel would be ideal for this, since even the table listing hits multiple APIs
08:29:02 <tsufiev> any other open discussion topics?
08:29:13 <robcresswell> :) Lets end here
08:29:16 <robcresswell> Thanks everyone
08:29:36 <robcresswell> #endmeeting