20:00:10 <xgerman> #startmeeting Octavia
20:00:10 <openstack> Meeting started Wed Jul 29 20:00:10 2015 UTC and is due to finish in 60 minutes.  The chair is xgerman. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:00:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
20:00:14 <blogan> hi
20:00:16 <openstack> The meeting name has been set to 'octavia'
20:00:16 <xgerman> #chair blogan
20:00:17 <Sherif> hi
20:00:17 <openstack> Current chairs: blogan xgerman
20:00:17 <minwang2> hi
20:00:21 <sbalukoff> Howdy, folks!
20:00:26 <bana_k> hi
20:00:36 <xgerman> #topic Announcements
20:00:36 <ajmiller> hi
20:00:47 <fnaval> hi
20:00:58 <xgerman> GSLB
20:01:06 <xgerman> #link https://www.mail-archive.com/openstack-dev@lists.openstack.org/msg58991.html
20:01:07 <rm_work> o/
20:01:10 <madhu_ak> hi
20:01:18 <xgerman> #link https://etherpad.openstack.org/p/GSLB_project_name_vote
20:01:24 <sballe> o/
20:01:36 <xgerman> and congratulate johnsom a new GSLB core!!
20:01:43 <bharath_> o/
20:01:45 <sballe> +1
20:01:48 <minwang2> +1
20:01:48 <fnaval> o/
20:01:49 <madhu_ak> +1
20:01:49 <Sherif> +1
20:02:02 <johnsom__> Thanks!
20:02:03 <blogan> congrats breh!
20:02:08 <fnaval> grats!
20:02:35 <xgerman> they are looking for drivers, aka people who help shape the servcie
20:03:16 <xgerman> V2 Panels
20:03:22 <blogan> can they call them puppet masters instead? so not to confuse with actual vendor drivers?
20:03:30 <sballe> lol
20:03:33 <xgerman> lol
20:03:41 <sbalukoff> Hah!
20:04:01 <xgerman> #link https://review.openstack.org/#/c/206757
20:04:12 <xgerman> dougwig got us a horizon-lbaas repo
20:04:31 <xgerman> and the OpenStack UX team some place to collaborate one esigns
20:04:57 <xgerman> #link https://openstack.invisionapp.com/d/main#/projects
20:05:10 * blogan applauds
20:05:11 <xgerman> to get accounts you need to talk to the people in the UX channel?
20:05:14 <sballe> ool
20:05:17 <sballe> cool
20:05:44 <xgerman> since the OpenStack UX PTL is from HP I went a different route ;-)
20:06:03 <blogan> aren't all PTLs from HP now?
20:06:51 <sballe> yeah soon HP == OpenStack ;-) and a little RAX
20:07:38 <xgerman> Liberty Schedule
20:07:40 <blogan> little ole rax, the little cloud that could
20:07:42 <xgerman> #link : https://wiki.openstack.org/wiki/Liberty_Release_Schedule
20:07:49 <johnsom__> Hahaha, little....
20:08:02 <xgerman> mestery said that they are cutting L2 today or tomorrow
20:08:15 <blogan> yeah if neutron-lbaas didn't throw a wrench in it
20:08:24 <sbalukoff> Oooh!
20:08:28 <sbalukoff> I smell an opportunity. XD
20:08:47 <blogan> its already been done
20:08:57 <sbalukoff> Sonofa...
20:09:12 <xgerman> so dougwig is being pulled off the boat right now?
20:09:38 <sballe> or out of the water. He is suppose to be fly fishing
20:09:40 <sbalukoff> What's this?
20:09:46 <sbalukoff> Oh! Ok.
20:10:29 <xgerman> any more announcements?
20:11:01 <xgerman> #topic Brief progress reports
20:12:07 <sbalukoff> Progress on the Neutron-LBaaS L7 stuff....   Just a few more things to do there (some of which is waiting on Evgeny)
20:12:21 <johnsom__> I did some benchmarking work on mysql using the repo module and found satisfactory performance.  I am now looking at the REST driver and getting it's networking up to date.  Plus lots of reviews to keep things rolling
20:12:37 <blogan> sbalukoff: have you tested it out yet? it doesn't actually allow sharing pools right now right?
20:12:47 <sbalukoff> It's been mentioned that the docs for this need updating. I have a review to update the L7 spec in neutron-specs. I also know the wiki needs updating. Were there any other places people wanted to see docs getting updated?
20:13:03 <crc32> I did some benchmarking with redis and it was great. But I checked in the MySQL solution instead.
20:13:07 <sbalukoff> blogan: You are correct. That's one of the things that needs fixing. :)
20:13:18 <blogan> sbalukoff: i think we need a devref at some point in tree, but thats out of your work's scope
20:13:27 <sballe> I have been looking at integration with anchor and trying to increase my number of reviews
20:13:37 <blogan> johnsom__: good to know
20:13:49 <rm_work> sballe: did you have a CR for any preliminary anchor work?
20:13:50 <johnsom__> 0.001806563186646 seconds per amphora @ 5000 updates under memory engine on a VMware devstack
20:14:02 <sbalukoff> johnsom__: When you say "satisfactory performance," what do you mean? Can you please share meaningful numbers?
20:14:14 <sbalukoff> Hah! Ok.
20:14:16 <minwang2> I did the health manager service part, almos done of my part
20:14:28 <sballe> I have been looking at what xgerman did in the rest driver
20:14:53 <sbalukoff> johnsom__: What does that equate to in updates/sec?
20:15:02 <Sherif> I have been working on the amphora flows and got the task flow working
20:15:03 <sballe> rm_work: Does that anser your question?
20:15:08 <blogan> johnsom__: does that mean you did 1000 inserts per second?
20:15:13 <rm_work> sballe: so no CR yet? kk
20:15:15 <blogan> sorry updates
20:15:25 <bharath_> I included the diskclean up part in the housekeeping.. Half way through writing unit tests for house keeping. I have been busy with other internal stuff
20:15:25 <xgerman> rm_work do you think we need one?
20:15:29 <rm_work> err
20:15:40 <rm_work> i mean, you'll want an Anchor implementation for CertGenerator
20:15:42 <rm_work> so yes?
20:15:43 <crc32> I get simular results when I use the repo.replace method instead of session.merge
20:15:51 <Sherif> a bug reported on taskflow for conditional flow at  https://bugs.launchpad.net/taskflow/+bug/1479466
20:15:52 <openstack> Launchpad bug 1479466 in taskflow "The needs to treat subflows as standalone entity" [Undecided,New]
20:15:52 * TrevorV feels sheepish... Forgot about the meeting
20:15:53 <uvirtbot> Launchpad bug 1479466 in taskflow "The needs  to treat subflows as standalone entity" [Undecided,New]
20:15:54 <uvirtbot> Launchpad bug 1479466 in taskflow "The needs  to treat subflows as standalone entity" [Undecided,New] https://launchpad.net/bugs/1479466
20:15:54 <rm_work> how else would you use anchor?
20:16:14 <sballe> rm_work: xgerman Let's sync-up after the meeting as a side bar
20:16:18 <rm_work> k
20:16:20 <blogan> tahts a lot of bots reporting
20:16:25 <xgerman> yeah
20:16:29 <madhu_ak> here are my updates
20:16:37 <Sherif> A workaround on conditional flow is implemented in: https://review.openstack.org/#/c/206252/
20:16:44 <madhu_ak> 1. After debugging the jenkins failures, scenario tests are passing in jenkins. Proposed a new patch at https://review.openstack.org/#/c/207196/. once the experimental job gives +1, it is goog to go i think
20:16:54 <madhu_ak> 2. Session persistence scenario test - Found some bugs and the details can be found in my comments at https://review.openstack.org/#/c/205626/
20:17:01 <madhu_ak> 3. ipv6 scenario tests: Has anybody worked on creating ipv6 lb and tested them before? I need some details to clarify in respect to send the requests to ipv6 lb for validation
20:17:03 <crc32> speaking of. Whats wrong with sonar?
20:17:07 <madhu_ak> 4. TLS Scenario tests: Rebased https://review.openstack.org/#/c/164828/ and its in WIP
20:17:13 <xgerman> johnsome
20:17:17 <xgerman> johnsom?
20:18:06 <xgerman> we will look into that later today
20:18:19 <sbalukoff> I think we lost him.
20:18:31 <xgerman> yeah, he was celebrating his new core-dom
20:18:32 <madhu_ak> no, here is it: johnsom__
20:18:43 <blogan> he must have been running his mysql tests :)
20:18:47 <madhu_ak> it is*
20:18:57 <sbalukoff> Haha
20:18:58 <johnsom__> sbalukoff 553.78.... amphora updates per second
20:19:17 <sbalukoff> crc32: What numbers did y'all need?
20:19:25 <johnsom__> blogan updates, inserts are slower, but our common case is updates
20:19:42 <blogan> johnsom__: yeah i didnt mean inserts, those are a one time thing really
20:20:09 <crc32> I get around 1000 amp heartbeats per second but anything abover that and the Queue starts to grow
20:20:41 <johnsom__> Yeah, sorry, I am remote today, so network challenged
20:21:23 <sbalukoff> crc32: Ok, my question was more about:  "What numbers does RAX actually need here?" With an eye toward: "Is looking into something other than MySQL actually a premature optimization?"
20:21:25 <blogan> i think we can get into the numbers later in the agenda for this
20:21:33 <blogan> unless we just knock it out now
20:22:01 <sbalukoff> Sorry, if you want to hit it later that's fine. Just seemed apropos since it was a progress update. :)
20:22:02 <xgerman> #topic Mysql and Redis storage for heartbeat data
20:22:02 <johnsom__> I think we all agree that Redis is end game, but good enough with mysql should work for now.  Ideally we would like to be able to use either.
20:22:04 <blogan> sbalukoff: it is, we will definitely need to go the redis route later witht eh density we expect
20:22:13 <xgerman> just switched agenda items :-)
20:22:25 <crc32> its not a premature optimization and were planning on interfacing it out so we can use redis. but for now MySQL is what were going with.
20:22:31 <blogan> johnsom__: yes agreed, coudl be another interface or just a conditional based on config values, but redis will eventually be used
20:22:59 <xgerman> yeah, I think we have sufficient abstraction with the mixin
20:23:10 <sbalukoff> Ok!
20:23:13 <blogan> btu using mysql for now will be fine, doing a redis implementation will be trivial i believe
20:23:21 <johnsom__> Personally I don't want to burn time getting a Redis setup for us under devstack.  My eye is on the Liberty goal
20:23:29 <xgerman> +1
20:23:37 <sballe> +1
20:23:39 <sbalukoff> Seems legit.
20:23:41 <sbalukoff> +1
20:23:42 <xgerman> I think reds would be more an operator thing
20:23:45 <minwang2> +1
20:23:50 <Sherif> +!
20:23:51 <Sherif> +1
20:23:55 <xgerman> sbalukoff  kght be contractually required to use DB2
20:24:04 <johnsom__> DB2 hahaha
20:24:05 <sbalukoff> I want to get L7 in before the liberty cut-off. :D
20:24:07 <sballe> lol
20:24:16 <sbalukoff> We'll see if that happens.
20:24:27 <sballe> somebody should give sbalukoff an IBM hat :-)
20:24:33 <blogan> well time was burned on a sonar job too, just saying :)
20:24:34 <sbalukoff> xgerman: Working on that driver now.
20:25:02 <xgerman> cool
20:25:05 <johnsom__> blogan yeah...  xgerman to comment?
20:25:22 <johnsom__> It's a handy tool actually.  I like it for reviews.
20:25:26 <xgerman> +!00
20:25:28 <blogan> when it works
20:25:30 <xgerman> +100
20:25:37 <xgerman> well, it runs on HP Cloud
20:25:41 <johnsom__> BTW, I will address the questions raised over lunch in open discussion
20:26:34 <blogan> i wasn't saying it wasn't handy btw
20:26:52 <xgerman> #decision use mysql for devstack; operators can pick different backend for prod
20:27:04 <blogan> anyway we're all in agreement on mysql first, then redis whenever its needed
20:27:14 <sballe> sounds good
20:27:18 <xgerman> +1
20:27:19 <sbalukoff> Yup.
20:27:32 <Sherif> +1
20:27:34 <minwang2> +2
20:27:34 <TrevorV> -2 (cuz I don't actually have that power)
20:27:43 <xgerman> #topic Octavia reference implementation in Liberty
20:28:18 <xgerman> #link https://etherpad.openstack.org/p/YVR-neutron-octavia
20:28:26 <xgerman> so I think we need to descope
20:28:55 <sbalukoff> Wow, lots to do there!
20:29:01 <sbalukoff> bana_k: If you're looking for something... ;)
20:29:09 <xgerman> yeah, and feature freeze is 8/31
20:29:42 <blogan> so much to do, so little time
20:29:46 <sballe> :-)
20:30:30 <blogan> need to get the octavia driver in good shape
20:30:33 <sbalukoff> Indeed!
20:30:36 <bana_k> def but needs some help though
20:30:47 <xgerman> no worries we are here to help :-)
20:30:52 <Sherif> So at end of liberty means Oct? right ?
20:31:00 <bana_k> :)
20:31:04 <blogan> Sherif: sep 1-3
20:31:08 <xgerman> Sherif, feature freeze is 8/31
20:31:14 <blogan> err what xgerman said
20:31:27 <sbalukoff> If it's not in before feature freeze, it doesn't get in. Only bugfixes after that until release.
20:31:45 <xgerman> yep + they have gotten pretty strict about that
20:31:50 <Sherif> Ok
20:32:03 <sbalukoff> You weren't planning on taking a vacation this summer, right? ;)
20:32:15 <blogan> dougwig apparently was
20:32:18 <xgerman> Sherif is an intern he has like two more months :-)
20:32:21 <sbalukoff> Slacker.
20:32:25 <xgerman> +!
20:32:26 <xgerman> +1
20:32:32 <blogan> interns are an expendable resource
20:32:37 <blogan> work them to the bone
20:32:38 <Sherif> Third summer in a row without a vacations
20:32:48 <sbalukoff> In case you actually do read the minutes afterward: We love you, dougwig!
20:32:59 <Sherif> misconception: intern == infinite energy :)
20:33:04 <xgerman> well, breaking the build just before vacation — hmmh
20:33:19 <sbalukoff> xgerman: that's how he shows he loves us!
20:33:22 <blogan> infinite energy only if consumed
20:33:35 <xgerman> Redbull? Works for dougwig
20:34:14 <xgerman> so let’s think that through is there stuff we can descope?
20:34:17 <Sherif> @blogan better managed than school session though.
20:34:51 <sballe> xgerman: any suggestions?
20:34:52 <Sherif> xgerman I have no idea . Sept - 12 is my last day
20:35:09 <xgerman> Minimum performance/scalability testing?
20:35:14 <xgerman> that can go :-)
20:35:15 <Sherif> I will fight fire with fire to get it done though
20:35:30 <sbalukoff> Wow, there's a lot of stuff on that list that isn't expendible. :P
20:36:08 <blogan> well i'd be fine with tempest tests just being in neutron-lbaas
20:36:13 <blogan> for now, its what the current namespace driver has
20:36:27 <xgerman> same here
20:36:32 <johnsom__> +1
20:36:37 <sballe> +1
20:36:45 <xgerman> documentation we might be able to add after deadline
20:36:46 <blogan> performance/scalability testing?
20:37:03 <sballe> I think it is premature so thta could go to
20:37:09 <blogan> same
20:37:11 <xgerman> +!
20:37:11 <sbalukoff> Right. The existing ref. impl. doesn't do any of that.
20:37:12 <xgerman> +1
20:37:17 <sbalukoff> So we shouldn't have to to replace it.
20:37:24 <sbalukoff> +1
20:37:29 <xgerman> well, the thinking was we shouldn’t be slower
20:37:39 <johnsom__> I think we have to have a gate setup, I don't think we can get out of that
20:37:40 <xgerman> so hence we need to measure :-)
20:37:46 <sbalukoff> How do you measure "slower" if you don't have any metrics to start from? ;)
20:37:55 <blogan> johnsom__: yes we need that
20:37:57 <sballe> :-)
20:38:08 <xgerman> sbalukoff agreed - let’s kill it
20:38:13 <blogan> sbalukoff: if its slower than infinite, then its better
20:38:14 <sballe> +1
20:38:22 <sbalukoff> Haha!
20:38:45 <xgerman> johnsom +1
20:38:58 <sballe> +1
20:38:58 <blogan> i actually dont see that much to do
20:39:08 <blogan> just some decisions that are already made
20:39:17 <blogan> like just removing namespace driver
20:39:22 <blogan> or at least pulling it out of tree
20:39:41 <blogan> do need to talk to packagers
20:39:58 <xgerman> yep
20:40:13 <sbalukoff> blogan: Will have it done by 4:30pm?
20:40:29 <blogan> i will have a nap done, yes
20:40:30 <xgerman> well, I think we need something working first?
20:40:46 <xgerman> after all Octavia only works with some major handholding
20:40:57 <blogan> okay so then lets get a more precise list down
20:40:59 <blogan> modify that etherpad?
20:41:11 <xgerman> cool with me
20:41:24 <sbalukoff> Yes, usability is a concern.
20:42:10 <sbalukoff> People trying to write code for the project have trouble figuring it out... so how can we expect to replace the reference driver if it's that much harder to get it going?
20:42:33 <sbalukoff> At least, we need "how to get started" documentation that works most of the time. :)
20:42:44 <sballe> +1
20:42:48 <xgerman> +1
20:42:50 <johnsom__> +1
20:42:52 <sbalukoff> But better defaults is probably a good idea.
20:42:55 <blogan> osmosis
20:43:00 <sbalukoff> (just sayin')
20:43:41 <johnsom__> sbalukoff Feel free to check in some different defaults.
20:43:51 <sbalukoff> johnsom__: Oh, I will. XD
20:44:03 <xgerman> but test them first :-)
20:44:17 <sbalukoff> I'm not blaming anyone here, yo-- I think y'all have made awesome progress!
20:45:57 <blogan> L7 in octavia is going to be just as much a pain as neutron-lbaas, needs pool sharing
20:46:05 <xgerman> yeah
20:46:06 <sbalukoff> Yes. Working on it.
20:46:15 <blogan> sbalukoff: working on it in octavia as well?
20:46:15 <sbalukoff> The code will end up being very similar.
20:46:21 <sbalukoff> Not in Octavia yet.
20:46:54 <xgerman> ok, I think we have sort of a list now on the etherpad
20:46:57 <sbalukoff> I wanted to get it going for the current ref. impl. first. But if I get stuck or delayed there, I might just reverse that and do it in Octavia first.
20:47:05 <sbalukoff> Opinions on this? should I do the Octavia work first?
20:47:06 <xgerman> +!
20:47:07 <xgerman> +1
20:47:09 <rm_work> WTB Single-Call Create
20:47:17 <xgerman> de-scoped?
20:47:21 <sbalukoff> rm_work: Too expensive for you!
20:47:25 <rm_work> heh
20:47:38 <blogan> sbalukoff: i'd feel better if namespace driver had it as a backup plan
20:47:49 <sbalukoff> Right.
20:47:59 <blogan> sbalukoff: sucks to basically redo code though
20:48:08 <sbalukoff> Well, and I'm also still working on getting roadmap docs updated. Haven't had as much time for that as I would have liked.
20:48:09 <blogan> sbalukoff: sucks even more to know the code may just be ripped out
20:48:26 <xgerman> nah, we will burn it
20:48:31 <sbalukoff> blogan: Yep, but there's no way around it until we replace that reference implementation.
20:48:44 <blogan> xgerman wants to watch the world burn
20:48:57 <xgerman> :-)
20:48:58 <sbalukoff> xgerman: Hah! I've got lighter fluid, if you've got a match.
20:49:16 <xgerman> next time I am in Seattle...
20:49:38 <sbalukoff> Seriously we should do a BBQ or something the next time y'all are in town.
20:49:45 <sbalukoff> I could deep fry a turkey or two...
20:50:00 <blogan> seattle bbq? is that like seaweed wrapped salmon?
20:50:05 <sbalukoff> Haha!
20:50:12 <sbalukoff> Well, "grilling"
20:50:26 <rm_work> heh
20:50:30 <sbalukoff> Around here we just call it all BBQ. I suspect to rub you Texans the wrong way.
20:50:32 <johnsom__> blogan hey, don't dis it until you try it
20:50:33 <rm_work> well, I'm up next month :P
20:50:38 <rm_work> for the ... 4th time this year
20:50:47 <xgerman> fun!
20:50:52 <rm_work> at least I am finally getting frequent flier miles properly
20:50:55 <sbalukoff> Anyway... back on topic.
20:50:57 <blogan> johnsom__: i tried it in pieces and it was delicious
20:51:05 <sbalukoff> I'm not seeing much else on this list that doesn't seem "essential."
20:51:10 <blogan> rm_work you should just move there, get out of texas!
20:51:15 <rm_work> heh
20:51:19 <rm_work> i wouldn't mind
20:51:31 <xgerman> RAX northwest office?
20:51:32 <blogan> should we get volunteers for these?
20:51:37 <blogan> L7 is obviously sbalukoff
20:51:43 <blogan> TLS is obviously ptoohill/pothole
20:51:47 <sbalukoff> Yep.
20:51:52 <xgerman> Yep
20:51:53 <blogan> I can work with doug on the octavia driver
20:51:59 <TrevorV> Can't offend me with the BBQ talk... I don't even care for BBQ like Texans like it.
20:52:05 <blogan> heartbeats/failover is in progress
20:52:06 <xgerman> sounds good
20:52:07 <crc32> I just want to go some where were its not 100 degrees.
20:52:08 <rm_work> i'll ... continue doing thorough reviews :P
20:52:08 <TrevorV> Except when pothole makes hot dogs on the grill outside... hell yeah
20:52:24 <rm_work> and making blogan sad
20:52:31 <pothole> :P
20:52:45 <rm_work> hopefully back on real code next month >_>
20:53:12 <sbalukoff> rm_work: Fight the good fight!
20:53:42 <blogan> who wants evil documetnation?
20:53:53 * johnsom__ steps back
20:54:04 * TrevorV steps behind min
20:54:04 <xgerman> do I need to volunteer people?
20:54:12 <xgerman> voluntell
20:54:37 <sbalukoff> Question: Does documentation need to be in before the code freeze?
20:54:46 <sbalukoff> It's not a new "feature" ;)
20:54:57 <xgerman> yeah, I think we can do that after
20:55:04 <xgerman> also bana_k didn’t pick something up yet
20:55:09 <sbalukoff> Well, I likes me some docs.
20:55:22 <sbalukoff> Yep: bana_k: Anything on that etherpad strike your fancy?
20:55:24 <xgerman> and he still needs to get his car towed, too
20:55:27 <bana_k> I dont know what to picK :)
20:55:50 <TrevorV> I can do the documentations stuffs, but it'll involve lots of talking to peeps I'm sure
20:55:52 <xgerman> towage can tell you all about our initiation ritual
20:55:56 <blogan> bana_k got his car towed?
20:55:57 <xgerman> towgan
20:56:01 <xgerman> not yet
20:56:41 <sbalukoff> bana_k: If you want to help me with the L7 stuff, I know that'd be appreciated.
20:56:52 <bana_k> Yes def :)
20:57:01 <blogan> isn't TrevorV helping out on the L7 stuff
20:57:05 <TrevorV> Also by documentation is that just spec-style docs, or prettification of startup stuffs and whatnot to get people rolling?
20:57:12 <sbalukoff> Also, everyone's code is going to need to be reviewed as we write this stuff.
20:57:28 <sbalukoff> TrevorV: The latter, mostly.
20:58:09 <sbalukoff> Specs are important, and we can back-fill them if the code is already written. But we're not going to be able to replace the reference implementation if nobody can figure out how to get Octavia working
20:58:45 <TrevorV> Alright, well, I can do the prettification stuffs too, just wondering what the idea was
20:58:53 <TrevorV> I made quite a few of our CLB 1.0 docs initially at Rax anyway
20:59:00 <TrevorV> So I can figure it out :D
20:59:01 <madhu_ak> I can pick one if you still needs people to play around with octavia stuffs after I am done with scenario tests
20:59:29 <blogan> should these docs be in the wiki or done as specs?
20:59:40 <xgerman> there was a third place
20:59:49 <xgerman> sc68cal?
20:59:54 <blogan> oh devref
20:59:56 <xgerman> network docs or so
21:00:03 <blogan> which we should kind of do
21:00:03 <xgerman> devref, yep
21:00:13 <blogan> ok devref
21:00:19 <blogan> would be good to get that going as well
21:00:23 <xgerman> well, times out
21:00:25 <blogan> TrevorV: look at neutron's devref
21:00:28 <xgerman> #endmeeting