15:00:19 <eglynn> #startmeeting ceilometer
15:00:19 <openstack> Meeting started Thu Jul 10 15:00:19 2014 UTC and is due to finish in 60 minutes.  The chair is eglynn. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:21 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:24 <openstack> The meeting name has been set to 'ceilometer'
15:00:27 <DinaBelova> o/
15:00:31 <ildikov> o/
15:00:46 <llu-laptop> o/
15:00:48 <idegtiarov> o/
15:00:58 <nealph> o/
15:01:13 <nsaje> hey guys o/
15:01:21 <cdent> hola
15:01:31 <KurtRao> o/
15:01:48 <eglynn> hey y'all!
15:01:53 <eglynn> #topic Juno-2 status
15:02:04 <eglynn> #link https://launchpad.net/ceilometer/+milestone/juno-2
15:02:16 <ildikov> my dispatcher BP for Gnocchi is under approval
15:02:30 <eglynn> ildikov: internal appro?
15:02:35 <jd__> o/
15:02:38 <ildikov> unfortunately it is the holiday season of the Swedish guys
15:02:46 <ildikov> eglynn: yes, of course...
15:03:03 <eglynn> LOL ... it's always holiday seaon in Sweden ;)
15:03:12 <jd__> internal appro? that sounds weird :)
15:03:15 <ildikov> so some delay is possible, but I'm really trying to push this through, only the last step is missing
15:03:28 * eglynn should up sticks and move to Sweden for all the holidays ...
15:03:43 <ildikov> eglynn: yes it is but, now really noone is available there...
15:04:06 <ildikov> eglynn: if you like that ugly weather, then totally worth to move ;)
15:04:22 <ildikov> jd__: thanks for sharing my pain ;)
15:04:24 <DinaBelova> dear core-team, that's j2 and it needs review :) https://blueprints.launchpad.net/ceilometer/+spec/mongodb-events-feature
15:04:31 <cdent> I reckon  progress on "consume notifications from ironic" has gone from slow to maybe "good"
15:04:43 <eglynn> jd__: yeah it does to such finely-tuned opensource ears as our own ;)
15:04:48 <ildikov> DinaBelova: sure, on my endless list, no worries
15:04:49 <eglynn> cdent: cool
15:04:55 * jd__ nods
15:04:56 <DinaBelova> ildikov, :D
15:05:05 <cdent> needs docs and validation of some of the names used
15:05:31 <eglynn> cdent: I've just updated
15:05:46 <cdent> ✓
15:05:59 <ildikov> eglynn: BTW the docco status is not that good, I guess I will not sleep much until the j2 deadline...
15:06:09 <eglynn> OK so at this point it would be great to have a heads-up if anyone's worried about slippage
15:06:13 <prad> just a reminder, if core reviewers can review my pending spec , I can try getting them partially in for juno-2
15:06:20 <prad> specs*
15:06:37 <eglynn> ildikov: can we divide and conquer on some of the docs work?
15:06:48 <eglynn> prad: yep, apols for being remiss on that
15:07:14 <ildikov> maybe if there is anyone who's willing to write some admin guide sections
15:07:17 <prad> np, just wanted those in asap so i can implement those soon
15:07:42 <eglynn> llu-laptop: do you think that any of the SNMP stuff is targetable to juno-2?
15:08:04 <DinaBelova> eglynn, may we set the https://bugs.launchpad.net/ceilometer/+bug/1330989 to some importance?
15:08:05 <uvirtbot> Launchpad bug 1330989 in ceilometer "[Hbase] Source filter doesn't work with _get_columns" [Undecided,In progress]
15:08:05 <llu-laptop> I'll try my best, but it's risky
15:08:21 <DinaBelova> eglynn, it's still undecided :(
15:08:25 <ildikov> eglynn: I'm not 100% sure how much it would be consistent, I mean the terms we use, how to describe things, etc, so I mean the best is to have the whole text from one author
15:08:36 <eglynn> llu-laptop: great, we can always bump it off to juno-3 if it doesn't make it
15:08:52 <ildikov> eglynn: anyway, I will scream, if I've reached that point
15:09:01 <eglynn> DinaBelova: what importance do you suggest?
15:09:14 <DinaBelova> eglynn, medium I guess
15:09:17 <eglynn> ildikov: what format is the docco in, docbook?
15:09:32 * eglynn makes it so ...
15:09:37 <llu-laptop> eglynn: yeah. i've started though I'm not sure if lsomla is satisfied with the spec or not
15:09:52 <ildikov> eglynn: I guess, XML-based
15:10:04 <eglynn> llu-laptop: lsmola is out on vacation for this week and next
15:10:10 <DinaBelova> eglynn, thank you sir :)
15:10:16 <ildikov> eglynn: we can discuss the status mid-next-week
15:10:21 <eglynn> llu-laptop: ... then he's off at EuroPython for another week
15:10:46 <eglynn> llu-laptop: ... so I think you should prolly just go ahead without his explicit appro
15:10:55 <llu-laptop> eglynn, ok i'll start first, assuming he won't come back with big suprise for me. :)
15:10:57 <eglynn> ildikov: cool
15:11:02 <ildikov> eglynn: and then decide if I need more hands or not
15:11:10 <eglynn> llu-laptop: thank you sir!
15:11:32 <eglynn> ildikov: cool, if needs be I can try helping get that over the line
15:11:32 <DinaBelova> ildikov, btw, let's do some polling-tempest fight tomorrow :)
15:11:44 <DinaBelova> ildikov, how do you look on that?
15:11:46 <ildikov> DinaBelova: ok, tomorrow is fine
15:11:50 <cdent> ildikov: I'd be intrested in helping some on the docs side of things if that's an option.
15:11:52 <DinaBelova> ildikov, cool, +1
15:12:02 <eglynn> cdent: cool, thank you sir!
15:12:10 * jd__ smiled 'cause he reads 'pillow-tempest fight'
15:12:14 <ildikov> DinaBelova: at what time is Sean available?
15:12:38 <DinaBelova> ildikov, about 6-7PM Moscow time he begins being here
15:12:52 <DinaBelova> ildikov, so smth like this :)
15:13:12 <ildikov> cdent: cool, thanks, I will see, how it could work, if I cannot make any good progress with it in the next very few days
15:13:16 <DinaBelova> ildikov, that's about 3-4UTC
15:13:27 <ildikov> DinaBelova: ok, cool, it's not that late :)
15:13:35 <DinaBelova> ildikov, indeed :)
15:13:49 <DinaBelova> sorry for the offtopic, please continue, folks :)
15:13:53 <ildikov> DinaBelova: ping e tomorrow if you think it's time to fight ;)
15:14:00 <DinaBelova> ildikov, np
15:14:03 <eglynn> so if anyone else has a juno BP that they're sitting on, please try to get the specs proposed by tmrw if at all possible
15:14:17 <eglynn> ... even just a rough placeholder
15:14:49 <eglynn> anything else WRT juno-2?
15:15:17 <cdent> just confirmation that the javelin stuff has seen no recent progress, will poke at people early next week
15:15:28 <eglynn> cdent: thanks
15:15:42 <eglynn> ok, moving on I guess ...
15:15:49 <eglynn> #topic TSDaaS/gnocchi status
15:15:56 <eglynn> jd__: anything to report?
15:16:16 <eglynn> jd__: I owe you a unit test I believe
15:16:19 <jd__> I've been implementing the aggregation stuff discussed last week during the sprint
15:16:34 <jd__> eglynn: yeah that would be really helpful, and thanks for testing btw
15:16:51 <jd__> I've been working today on some new shiny features in the indexer
15:17:00 <jd__> so things are progressing :)
15:17:05 <ildikov> BTW, I couldn't figure out why MySQL is not working on the gate
15:17:15 <jd__> and I'm waiting for ildikov and her internal team :D wrt ceilometer integration
15:17:28 <jd__> ildikov: I missed that I guess?
15:17:31 <eglynn> jd__:  ... up to now I've just been using a little shell script to hit the API, need to translate that to an in-tree test
15:17:33 * DinaBelova working on the opentsdb driver for gnocchi. I believe it'll help to find what shall we do with the api and common classes to be abstract and good enough for three different backends :)
15:17:44 <jd__> cool DinaBelova
15:17:50 <ildikov> jd__: you can tell it my managers too, if you wish ;)
15:18:00 <DinaBelova> ildikov, lol :D
15:18:13 <eglynn> DinaBelova: what do the data retention semantics look like for opentsdb?
15:18:31 <DinaBelova> eglynn, /me still in progress here
15:18:39 <ildikov> jd__: with the socket included in the URL it cannot connect, otherwise it has authentication problems and locally I have a third issue with the tests with MySQL
15:18:40 <eglynn> DinaBelova: ... i.e. the equivalent of the stuff we were talking to pauldix about last week for influx
15:18:46 <eglynn> DinaBelova: ... a-ha, cool
15:18:53 <DinaBelova> eglynn, yes, I keep that in mind
15:19:06 <ildikov> jd__: more details: https://review.openstack.org/#/c/104966/
15:19:11 <jd__> ildikov: I though the gate was passing? I'll take a look
15:19:44 <ildikov> jd__: it passed, because all the tests were skipped, because there wasn't any working MySQL connection
15:20:02 <eglynn> jd__: just for simplicity, would it be easy to have the tests run in-tree against sqlite for the indexer driver?
15:20:19 <ildikov> jd__: anyway, I just wanted to have more eyes on it, as I got distracted much this week, so I didn't have enough time to deal with it
15:20:28 <jd__> eglynn: sqlite does not support foreignkey so it's not working with Gnocchi as we need that
15:20:42 <eglynn> jd__: a-ha k, scratch that so
15:20:44 <jd__> so I just dropped the support for it since it's not really worth it
15:21:11 <jd__> ildikov: I just saw that, my bad I sucked missing that
15:21:13 <ildikov> jd__: with MySQL I got AttributeError for with_for_update...
15:21:22 <cdent> sqlite3.something has foreign keys...
15:21:24 <jd__> ildikov: ok I'll fix that
15:21:30 <jd__> cdent: so I heard
15:21:45 <jd__> cdent: but still not really convinced my effort is going onto that direction anytime soon :D
15:21:45 <ildikov> jd__: np, if you have time, please check, I will be too busy to fix it this week for sure
15:21:53 <ildikov> jd__: cool, thanks
15:22:52 <eglynn> anything else on gnocchi?
15:23:09 <DinaBelova> eglynn, I guess that's it :)
15:23:18 <eglynn> #topic Mid-cycle wrap-up
15:23:43 <eglynn> it was fun and productive I think :)
15:23:52 <eglynn> jd__: thanks for the hospitality!
15:24:02 <DinaBelova> eglynn, and that amazing french bakery... mmmm
15:24:02 <jd__> you're welcome, it was fun having you guys!
15:24:11 <ildikov> eglynn: +1
15:24:21 <ildikov> I want more mid-cycles ;)
15:24:32 <DinaBelova> ildikov, hehe :D
15:24:32 <cdent> mid-mid-cycle
15:24:42 <jd__> 1/6th-cycle every month
15:24:58 <eglynn> yeah, do folks feel it's something we should aim to do again for Kilo/Kyoto?
15:24:58 <ildikov> jd__: thanks much, everything was really good!
15:24:58 <llu-laptop> all logs are at https://etherpad.openstack.org/p/paris-July2014-ceilometer-sprint?
15:25:07 <jd__> eglynn: likely
15:25:16 <eglynn> llu-laptop: yeah, they're a little rough TBH
15:25:19 <DinaBelova> llu-laptop, and in the links published there
15:25:46 <eglynn> llu-laptop: ... may not be very coherent, need to be read *cough* forensically
15:27:02 <eglynn> k, Paris was fun, but back to reality ... :)
15:27:08 <eglynn> #topic "The continuing saga of branchless Tempest"
15:27:36 <vrovachev> ohhh
15:27:49 <ildikov> eglynn: I've just read your yesterday's discussion with Joe about the notifications
15:27:51 <eglynn> so this discussion started up with vrovachev's featureful backport of the cinder snapshot notification support
15:28:18 <eglynn> I then raised it at the PTLs meeting which led to a long inconclusive discussion
15:28:25 <eglynn> ... which we then punted to the ML
15:28:28 <eglynn> http://lists.openstack.org/pipermail/openstack-dev/2014-July/039733.html
15:28:48 <eglynn> and as ildikov says, continued on IRC with jogo last night
15:29:13 <eglynn> so there are several interleaved threads of thought and lines of reasoning
15:29:19 <DinaBelova> this mailing thread is growing with each minute :(
15:29:35 <ildikov> eglynn: is there any chance to have a freeze and then very slowly move to one direction with notifications?
15:29:51 <ildikov> eglynn: just like what we were talking about on the mid-cycle
15:29:59 <eglynn> but in the short term ... seems that Matt is accepting of a test-skipping decorator based on static (i.e. non-discoverable) tempest config
15:30:13 <eglynn> ildikov: a freeze?
15:30:35 <vrovachev> I want to eliminate -1 in Jenkins using api ceilometer(capabilities:storage:production_ready)
15:31:15 <eglynn> vrovachev: "eliminate -1 in Jenkins" ==> is that causing CI failures?
15:31:34 <vrovachev> eglynn: yes
15:31:48 <DinaBelova> eglynn, in grenade job afair
15:31:57 <ildikov> eglynn: I just meant that somehow we need to deal with the changing style of notifications and I guess there was a point in your discussion to have a state and versioning, etc
15:32:00 <vrovachev> DinaBelova: yes :)
15:32:23 <eglynn> vrovachev: can you explain the background? ... what changes do we need to avoid that failure?
15:33:52 <eglynn> ildikov: yep, if the other projects step up to actually versioning notifications, then that change would impact on consumers such as ourselves and StackTach
15:33:57 <vrovachev> grenade job runned tempest in master branch for stable icehouse devstack. I can to add skip test, if get API request to apabilities:storage:production_ready is False
15:34:05 <eglynn> ildikov: ... but it would be to our benefit, right?
15:34:44 <eglynn> vrovachev: a-ha, so the same skipping logic as you already put into the main devstack job?
15:34:51 <ildikov> eglynn: yeap, smth like that
15:35:04 <vrovachev> eglynn: yes
15:35:19 <eglynn> ildikov: k, I just didn't get the freeze reference ... i.e. what was being frozen
15:35:22 <DinaBelova> eglynn, yes, smth like that - that was fixed by the branches in devstack, but here it's an other situation
15:35:34 <DinaBelova> a little bit
15:35:41 <eglynn> vrovachev, DinaBelova: ok
15:35:43 <ildikov> eglynn: I still think that it would be much easier to have a some kind of stardized format in the *long-term* future, but still, we have to start at somewhere to even deal with this issue
15:35:57 <DinaBelova> although, it might be fixed with the API specs endpoint I've done for the ceilo that time
15:36:02 <eglynn> vrovachev: ... so on the cinder tests
15:36:05 <cdent> ildikov: that's basically the same as what I just posted to the thread
15:36:26 <vrovachev> eglynn: yes, on snapshot notification
15:36:30 <eglynn> vrovachev: ... does it Matt's proposed "@test.consumed_notifications('volumes', 'snapshot.exists', 'snapshot.create.*', ...)" style of decorator make sense to you?
15:36:44 <vrovachev> on volume notification all ok :)
15:36:58 <ildikov> cdent: I'm trying to give this hint to people around for a few months now..., unfortunately it's not that easy as it sounds... :(
15:37:00 <eglynn> vrovachev: ... i.e. as described in http://lists.openstack.org/pipermail/openstack-dev/2014-July/039819.html
15:37:06 <DinaBelova> eglynn, I guess it'll look nice - esp. if QA is ok with it
15:37:19 <ildikov> DinaBelova: +1
15:37:35 <eglynn> DinaBelova: it sounds like they are, as Matt has given his blessings
15:37:43 <DinaBelova> eglynn, yes, indeed
15:38:02 <eglynn> vrovachev: so is that something you could implement so as to avoid the need for the cinder snapshot backport
15:38:04 <eglynn> ?
15:38:29 <DinaBelova> I think that'll work ok, although I'm still interested how to implement that in tempest, as I'm not the specialist. vrovachev - will it be ok for you?
15:39:35 <vrovachev> eglynn: I will discuss code style (implemented decorator or skip in test) with Sean and Matthew. Yes, this can avoid the need for the cinder snapshot backport
15:39:49 <eglynn> vrovachev: excellent!
15:40:01 <DinaBelova> yay :)
15:40:13 <vrovachev> :)
15:40:38 <eglynn> vrovachev: ... so with that skipping approach agreed, is it realistic do you think to get the stalled tempest tests landed for juno-2?
15:41:04 <vrovachev> for cinder  and swift - juno 2
15:41:10 <vrovachev> for glance - juno 1
15:41:37 <eglynn> vrovachev: juno 1 ... do you mean, already landed?
15:41:53 <DinaBelova> eglynn, also scenario tests (events, alarms) are in progress for j2
15:42:05 <vrovachev> oh, no, for nova - juno 1 :)
15:42:21 <DinaBelova> eglynn, for nova they're landed
15:43:01 <vrovachev> for glance juni 2 also
15:43:10 <vrovachev> juno_
15:43:33 <eglynn> DinaBelova, vrovachev: excellent, so it would really geat to be able to declare victory on the TC gap for tempest in juno-2 :)
15:43:46 <DinaBelova> eglynn, /we are in progress :D
15:43:51 <eglynn> the other thing WRT testing is ...
15:44:00 <eglynn> did folks read Sean's magnum opus on that thread? ... http://lists.openstack.org/pipermail/openstack-dev/2014-July/039733.html
15:44:33 <DinaBelova> eglynn, indeed :D
15:44:55 <DinaBelova> it is a nice story :)
15:45:07 <eglynn> particularly the "New Thinking about our validation layers" section
15:45:24 <vrovachev> oh, yes...
15:45:39 <eglynn> ... so it seems the wind is blowing somewhat towards in-tree functional tests outside of tempest
15:45:46 <cdent> thank god for that
15:45:52 <eglynn> ... just wondered what folks thought of that idea?
15:46:34 <eglynn> cdent approves?
15:46:50 <DinaBelova> eglynn, I just wonder how difficult it'll be to be done
15:47:22 <DinaBelova> as tempest is now big, but more or less known thing - I wonder how this might look like outside the tempest
15:47:23 <cdent> In general I prefer maximizing tests in-tree
15:47:37 <vrovachev> 'in-tree functional tests outside of tempest' this Sean's idea , I'm not thinking, good or bad this
15:47:55 <eglynn> cdent: in ceilo's case, would it end up being effectively a mini-Tempest?
15:48:27 <eglynn> cdent: ... i.e. can we really system-test ceilo in isolation in a meaningful way?
15:48:37 <cdent> certainly some of it
15:48:38 <vrovachev> I think, functional and integrative test outside the tempest it's bad idea
15:48:58 <cdent> we already require a light mongo for some tests, yes?
15:48:58 <vrovachev> _integrated tests
15:49:11 <cdent> and things like wsgi-intercept can be used for more end2end api tests
15:50:00 <eglynn> so I'm a bit nervous that we'd end up with a bunch of fragmented, diverging, duplicated efforts to build mini-Tempests for each project
15:50:12 <eglynn> ... to do stuff that maxi-Tempest could do prior to the advent of branchless
15:50:29 <DinaBelova> eglynn, +1
15:50:38 <cdent> well presumably maxi-tempest is inter-project and ceilo's mini-tempest is intra-project?
15:50:47 <vrovachev> check the notifications in tempest very well, because we always know not lost any connection with any project.
15:51:00 <vrovachev> for pollsters also
15:51:11 <DinaBelova> vrovachev, pollsters is other story :D
15:51:20 <vrovachev> oh...
15:51:21 <eglynn> cdent: yep, and mini-tempest is whitebox, versus maxi-tempest "kinda" blackbox
15:51:23 <DinaBelova> we need to persuade that it's needed :D
15:52:54 <eglynn> ok, so I'd encourage folks with strong opinions either one way or the other to chime in on that thread
15:53:33 <vrovachev> eglynn: of course
15:53:49 <eglynn> ... I suspect we'll end up with much discussion about this in-tree functional testing idea next time we're in Paris
15:54:05 <eglynn> (if it doesn't come to the fore even earlier than that)
15:54:17 <eglynn> ok, let's move on ...
15:54:25 <cdent> I'm not undertanding why it is a bad thing on any dimension? Can someone explain that to me out of band?
15:54:33 <eglynn> #topic Eventlet to asynchio switch-over
15:55:21 <eglynn> just a heads up something that markmc, victor stinner & flaper87 discussed at length at the mid-cycle last week
15:55:27 <eglynn> http://lists.openstack.org/pipermail/openstack-dev/2014-July/039291.html
15:55:40 <llu-laptop> when is the switch-over required to complete? juno?
15:56:11 <eglynn> tl;dr: deprecate eventlet in baby steps, replace piecewise with backport of py3 asyncio (trollius), use ceilo as the guinea pig
15:56:35 <eglynn> llu-laptop: ... well I was thinking it would have to be a multi-cycle effort
15:56:45 <eglynn> llu-laptop: ... maybe the first steps by end of juno
15:57:10 <llu-laptop> eglynn: relieved to hear this.
15:57:31 <eglynn> ... but there's a few potential pifalls raised on that thread, sql-a coexistence with asyncio for example
15:57:56 <eglynn> ... so still a topic of live discussion, worth keeping at eye on
15:58:11 <cdent> there's another related thread on the sql-a
15:58:36 * eglynn doesn't pretend to understand all the subtleties being raised on the ML, but good have it aired in any case and addressed by the domain experts
15:58:47 <eglynn> k, running out of time here
15:58:52 <llu-laptop> cdent: what's he link to the sql-a thread?
15:58:58 <cdent> one sec
15:59:30 <eglynn> #topic open discussion
15:59:46 <eglynn> ... while cdent is digging for that link
16:00:03 <cdent> http://lists.openstack.org/pipermail/openstack-dev/2014-July/039716.html
16:00:09 <llu-laptop> cdent: thx
16:00:29 <cdent> it's only part of the puzzle, but some people are like "zomg, nightmare" and others are "NBD"
16:01:15 <eglynn> NBD == no big deal?
16:01:43 <eglynn> presumably so
16:01:56 <eglynn> if anyone's interested, /me was telling lies about Juno again
16:02:01 <eglynn> ... https://plus.google.com/events/c6e8vjjn8klrf78ruhkr95j4tas
16:02:14 <eglynn> any got anything else?
16:02:22 <eglynn> ... we're at minus 2 mins ;)
16:02:31 <eglynn> k, let's call it a wrap
16:02:38 <eglynn> thanks folks!
16:02:42 <eglynn> #endmeeting