20:00:12 <johnsom> #startmeeting Octavia
20:00:12 <openstack> Meeting started Wed Oct 12 20:00:12 2016 UTC and is due to finish in 60 minutes.  The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:00:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
20:00:15 <openstack> The meeting name has been set to 'octavia'
20:00:17 <sbalukoff> Howdy, folks!
20:00:25 <diltram> hello all
20:00:29 <johnsom> Hi there
20:00:58 <johnsom> Ok, getting started
20:00:59 <johnsom> #topic Announcements
20:01:29 <johnsom> I have proposed a new core on the mailing list.  If you are a current core and have not yet voted, please do so.
20:02:06 <johnsom> Dougwig and I will be hosting a fishbowl session at the design summit
20:02:15 <johnsom> #link https://www.openstack.org/summit/barcelona-2016/summit-schedule/global-search?t=Neutron%3A+LBaaS+retrospective+and+next+steps
20:02:41 <johnsom> Please attend if you are coming to the summit.  I expect the bulk of the discussion to be about the merge work.
20:03:00 <johnsom> Please add ideas to the etherpad if there is a topic you would like to talk about.
20:03:39 <johnsom> Also, if you didn't see my post yesterday, I'm setting up grafana for Octavia so we can have a dashboard on job failure rates.
20:03:45 <sbalukoff> Yay!
20:03:46 <johnsom> #link https://review.openstack.org/#/c/385067/
20:04:09 <diltram> great
20:04:18 <johnsom> At a minimum it will make my life easier by having a simple place to look at our periodic job health.
20:04:35 <johnsom> Any other announcements today?
20:04:47 <xgerman> o/
20:04:56 <sbalukoff> None from me. :)
20:05:03 <johnsom> #topic Brief progress reports / bugs needing review
20:05:10 <johnsom> Ok, stuff going on...
20:05:23 <sbalukoff> Dustin and I have been spending some time working on some C10k optimizations.
20:05:27 <johnsom> I have been working through the review backlog.  We are making good progress there
20:05:56 <diltram> #link https://review.openstack.org/358158
20:06:06 <sbalukoff> In our tests, our current amphora image falls over with a standard HTTP listener getting 15k simultaneous connections. Though, of course you have to raise the connection limit on the listener.
20:06:07 <diltram> here is my review for Octavia API
20:06:53 <diltram> nice results
20:06:55 <sbalukoff> But it also looks like a 2 or 3GB image will handle something a lot larger than that (we hit 30k connections no problem with 2GB of RAM).
20:06:56 <johnsom> sbalukoff Ugh, that is lower than I expected
20:07:22 <diltram> what you're using to measure that?
20:07:24 <johnsom> Ah, so we are hitting a RAM issue?
20:07:26 <dougwig> o/, sorry i'm late.
20:07:42 <johnsom> I was expecting ~40k out of our lightly tuned amp
20:08:02 <sbalukoff> diltram: Just apache bench for now.
20:08:25 <sbalukoff> diltram: And Dustin's go-based webserver which keeps track of conection counts.
20:08:57 <diltram> ok, sounds good
20:09:20 <sbalukoff> Speaking of which:
20:09:27 <sbalukoff> #link: https://review.openstack.org/#/c/382067/
20:09:52 <dougwig> i've been eagerly awaiting a wall of green on that one.
20:10:12 <johnsom> Haha, poor dustin.  He has hit a number of snags on that
20:10:39 <sbalukoff> Heh! Well, our C10k tests uncovered most of them. So, I think we're pretty close to having something really useful there, eh.
20:10:46 <johnsom> Hey, it looks like that one might pass
20:11:30 <sbalukoff> Do we know how beefy the gate boxes are?
20:11:42 <johnsom> Not very.  The OSIC ones are the best
20:11:50 <dougwig> 8GB ram + swap, 30gb disk, 2-4 cores.
20:11:57 <sbalukoff> We were thinking of adding a C10k scenario test, but you obviously need to have instances that can handle the load.
20:12:15 <sbalukoff> Yeah... don't think we can do an effective test with that. :/
20:12:27 <sbalukoff> Ah well... we can at least add a "minimal" one.
20:12:31 <dougwig> you can put it on the a10 ci.  just submit a PR to its project-config.  i can add bigger nodes.
20:12:44 <sbalukoff> Which yes, is totally an oxymoron. ;)
20:12:54 <sbalukoff> dougwig: Ok! We'll keep that in mind, eh!
20:12:58 <johnsom> Yeah, my experience is the performance varies greatly between the hosts.  The A10 ci would be cool.
20:13:13 <johnsom> Any additional tuning patches coming?
20:13:30 <sbalukoff> johnsom: A couple minor ones probably.
20:13:40 <johnsom> Ok, cool.
20:13:44 <sbalukoff> And probably a document describing underlying hardware node tuning.
20:13:59 <sbalukoff> Because... well, there are limitations that can only be solved there, too.
20:14:01 <diltram> great that this will be publicly available
20:14:04 <johnsom> Yeah, there is a lot that can be done there.
20:14:46 <johnsom> Anyway, my plan is to continue doing reviews and then dig into some merge work
20:14:50 <sbalukoff> These are the *extremely* rough notes we've been keeping so far: https://etherpad.openstack.org/p/octavia-c10k-notes
20:14:54 <sbalukoff> Er...
20:15:03 <sbalukoff> #link: https://etherpad.openstack.org/p/octavia-c10k-notes
20:15:53 <johnsom> Any other updates to chat about?
20:16:10 <johnsom> #topic PTG planning / space
20:16:22 <johnsom> There is a bunch of mailing list traffic about the PTG.
20:16:48 <johnsom> From the responses I'm seeing we are small fish in a not-so-organized pond.
20:17:04 <diltram> true
20:17:18 <johnsom> So, I am asking Armando if we can crash the neutron space.
20:17:26 <dougwig> the gist of the ptg is that it's a week long, with 2 days of "cross project".  so we'd be down to 3 days, and given the increased attendance, i doubt code sprinting would be as viable as we normally strive for.
20:17:33 <johnsom> Instead of battle to get some of our own.
20:18:11 <johnsom> Yep.  It's not clear to me how useful this will be for us, but I am trying to plan ahead.
20:18:16 <dougwig> i wonder if the ptg is an upgrade from a standalone midcycle for us.
20:18:24 <johnsom> downgrade?
20:18:36 <dougwig> that'd be the other half of my wondering.
20:18:40 <sbalukoff> Hah! "upgrade"
20:18:54 <diltram> we will see
20:19:04 <johnsom> I feel that our mid-cycles have been productive.  This seems a bit like a zoo with distractions
20:19:06 <diltram> everything dependence of us
20:19:20 <dougwig> it feels like a cross project win, and a project lose, and an overall travel budget fail.  but i'm a pessimist.
20:19:27 <johnsom> Anyway, that is my plan.  Feedback welcome.
20:19:29 <sbalukoff> Yeah, I've been happier with our mid-cycles than the summits, FWIW.
20:19:41 <sbalukoff> Sounds like a good plan to me.
20:19:44 <diltram> we can go right now and if it's not gonna work we can organize our own mid-cycle's as previously
20:20:00 <dougwig> aye, it is the first one.
20:20:01 <johnsom> Yeah, I'm willing to give it a shot
20:20:11 <diltram> agree
20:20:31 <johnsom> #topic Translation for Octavia
20:20:47 <johnsom> Time to put dougwig in the spotlight
20:21:02 <dougwig> oh fuck.
20:21:08 <sbalukoff> Haha!
20:21:20 <johnsom> #topic Debian package build job for Octavia
20:21:29 <sbalukoff> Wait, what?
20:21:31 <johnsom> And I expect the same here so....
20:21:32 <dougwig> now you're just rubbing it in.
20:21:36 <sbalukoff> Oh!
20:21:38 <johnsom> #topic Open Discussion
20:21:42 <sbalukoff> Got it.
20:21:54 <johnsom> The two research topics
20:21:56 <sbalukoff> So that was our finger-waggling at dougwig?
20:22:03 <johnsom> Yep
20:22:04 <dougwig> yes.
20:22:04 <sbalukoff> But he's not even bleeding!
20:22:19 <dougwig> my soul is bleeding.
20:22:20 <johnsom> That starts after it has been a month....
20:22:29 <johnsom> HA
20:22:31 <johnsom> I bet
20:22:31 <dougwig> whoa, you might want to consider phrasing.
20:22:42 <johnsom> Oye
20:22:46 <johnsom> sorry
20:22:46 <sbalukoff> Haha!
20:22:55 <sbalukoff> That was pretty terrible.
20:22:58 <dougwig> else i'm gonna nickname you trump.
20:22:58 <johnsom> I am slow today.
20:23:15 <sbalukoff> Oh! Geez-- I thought you were referring to me not being here in a month.
20:23:17 <johnsom> You went there, I didn't  (obviously)
20:23:25 <sbalukoff> Because why else would dougwig's heart bleed?
20:23:50 <johnsom> ANYWAY, any other discussion items today?
20:24:14 <sbalukoff> Yay for short-ish meetings?
20:24:23 <sbalukoff> Thanks y'all!
20:24:23 <johnsom> Thanks for the continued reviews/help/patches/etc.
20:24:23 <dougwig> i think we killed all further discussion.
20:24:33 <johnsom> sigh
20:24:49 <johnsom> #endmeeting