15:16:15 <eglynn> #startmeeting ceilometer
15:16:16 <openstack> Meeting started Thu Feb 12 15:16:15 2015 UTC and is due to finish in 60 minutes.  The chair is eglynn. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:16:17 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:16:19 <openstack> The meeting name has been set to 'ceilometer'
15:16:19 <eglynn> #topic Kilo-3 status
15:16:26 <sileht> o/
15:16:38 <_elena_> o/
15:16:44 <fabiog> hi
15:16:49 <llu-laptop> o/
15:16:56 <idegtiarov> o/
15:17:11 <ildikov> o/
15:17:22 <gordc> o/'=
15:17:28 <gordc> ...
15:17:59 <linuxhermit> o/
15:18:22 <cdent> o/
15:19:04 <eglynn> #link https://launchpad.net/ceilometer/+milestone/kilo-3
15:19:37 <eglynn> ^^^ does that capture everything?
15:19:51 <eglynn> it includes a few holdovers from kilo-2
15:20:29 <cdent> the functional testing spec (which I think is still pending) probably will go in there?
15:20:48 <eglynn> cdent: yep
15:20:59 <llu-laptop> eglynn: would you please add https://blueprints.launchpad.net/ceilometer/+spec/power-thermal-data into kilo-3?
15:21:02 <gordc> looks ok to me. i think me ityaptin will add one for event transformer/bracketing related work.
15:21:19 <eglynn> llu-laptop: sure ... is Edwin working on it?
15:21:37 <llu-laptop> eglynn: I think so
15:21:52 <eglynn> llu-laptop: cool if he has the bandwidth to work on it
15:21:58 <eglynn> so heads up on the dates
15:22:05 <eglynn> kilo-3 is due on March 19th
15:22:21 <eglynn> but we have a feature proposal freeze two weeks prior
15:22:43 <eglynn> that means initial version of patches need to be proposed for Mar 5th
15:23:02 <eglynn> ... we aimed for that last time too, didn't quite make it ;)
15:23:28 <eglynn> it basically reduces the need for last minute feature freeze exceptions
15:23:48 <ityaptin> o/
15:24:02 <ildikov> and the last minute gate issues too hopefully
15:25:06 <eglynn> yeap
15:25:34 <eglynn> let's move on (since I burned the first 15 mins)
15:26:35 <eglynn> #topic gnocchi status
15:27:45 <eglynn> so as I said before, we need owners for everything in http://bit.ly/gnocchi-ceilo-k3-tasks
15:27:46 <ityaptin> I write to paper all storage driver features by last 2 months for implement it in influxdb and opentsdb CRs
15:27:56 <eglynn> thanks to all who volunteered
15:28:09 <eglynn> blueprints can now be filed for gnocchi
15:28:18 <eglynn> https://blueprints.launchpad.net/gnocchi
15:29:37 <eglynn> jd__: IIRC the TC confirmed on Tuesday evening that they're open for business in terms of accepting big tent applications
15:30:04 <jd__> cool that's good news
15:30:14 <jd__> I've made some progress on the gnocchi client – working on openstacksdk
15:30:19 <jd__> and I'm still working on that
15:30:36 <eglynn> jd__: yeah, I saw the patch for capabilities, nice
15:30:37 <jd__> I'm also working on adding to openstackclient
15:30:49 <jd__> final goal is to have "openstack metric create ..."
15:31:23 <jd__> other than that we're on a good road I think :)
15:32:24 <sileht> I got accepted new resources for heat and I finished the setup of gate/devstack, we have now 1 functional test :p
15:33:21 <eglynn> sileht: yeah, good getting the buy in on that for heat
15:33:27 <eglynn> *from heat
15:34:07 <eglynn> move on?
15:34:14 <eglynn> #topic swift middleware support in juno tests
15:34:29 <cdent> s'all good
15:34:30 <eglynn> gordc the floor is your's  ... ^^^
15:35:17 <gordc> sorry wasn't paying attention...
15:35:57 <eglynn> gordc: no worries ... so do you want to expand on that?
15:36:00 <gordc> so basically this is to discuss ongoing issue regarding swift middleware and it's dependecies
15:36:16 <eglynn> gordc: I'm not sure that backporting removal would be kosher
15:36:31 <eglynn> a-ha that's not what you're suggesting
15:36:40 <eglynn> just backport test changes, amiright?
15:36:43 <gordc> we have the new middleware in a separate library... but it won't work against juno since the listeners aren't in juno
15:37:06 <gordc> so i guess the question is do we want to backport the listeners to juno
15:37:09 <eglynn> a-ha, got it
15:37:18 <cdent> I'm not sure it is worth the effort
15:37:26 <cdent> Last night's failures were a red herring, not ceilo's fault.
15:37:29 <gordc> that or we continue as is... and handle dep issues as they come.
15:37:32 <gordc> cdent: yeah.
15:37:32 <eglynn> so strictly speaking that would not be a backportable thing according to the upstream criteria
15:37:52 <eglynn> https://wiki.openstack.org/wiki/StableBranch#Appropriate_Fixes
15:37:54 <gordc> just like to reach a consensus going forward i guess.
15:38:06 <cdent> (I added some additional notes to the bug to reflect last night's research)
15:38:23 <eglynn> New features == completely verboten
15:38:26 <eglynn> however
15:38:35 <eglynn> I can see certain distros backporting
15:38:43 <gordc> or we turn off swift tests as a third option... but i guess we're all leaning to just tracking issues as they come
15:39:28 <eglynn> gordc: turn off grenade tests, or?
15:39:33 <ildikov> gordc: I'm leaning
15:39:37 <cdent> I think just leave things as they are, see what happens.
15:39:53 <eglynn> cdent: the Irish solution :)
15:40:03 <ildikov> eglynn: LOL :)
15:40:04 <cdent> My mom was born there...
15:40:06 <gordc> lol... programmers solution.
15:40:09 <linuxhermit> :)
15:40:14 <llu-laptop> speaking of the gate test for old version, why do have have gate-tempest-dsvm-neutron-src-python-ceilometerclient-icehouse for ceilometerclient?
15:40:24 * eglynn hands cdent honourary citizenship :)
15:40:43 * cdent accepts his bottle
15:40:48 <linuxhermit> haha
15:40:54 <eglynn> LOL :)
15:41:19 <llu-laptop> I saw that task overwrite ceiloclient's requirement with icehouse version, and test the latest ceiloclient code on the overwrote requirements, that doesn't seems right to me
15:41:19 <gordc> i think we can close off i guess... the capping of requirements in juno should probably minimise any dep issues anyways
15:41:20 <eglynn> llu-laptop: good question, I don't know ... do any of the other python-*clients have similar
15:41:27 <eglynn> gordc: ok
15:41:50 <eglynn> #topic the real world/ready for primetime threads
15:42:03 <eglynn> cdent: thanks for bringing this one up
15:42:12 <eglynn> it does not make for comfortable reading
15:42:27 <cdent> Yeah, I added that because it's just sort of embarassing and I think we need to be more aggressive about responding and reacting in an active and productive way, not just ignoring it.
15:42:29 <eglynn> I have not reply as the OP is specifically looking for operational experiences
15:42:36 <gordc> i read the one which mentioned there's a doc that says mysql is the recommended backend.
15:42:40 <eglynn> (not rebuttal from us)
15:42:47 <cdent> Things are much better now since Juno, but I'm not really sure how much better.
15:42:50 <gordc> we need to hunt down that doc and burn it to the ground
15:43:29 <gordc> cdent: i think mongo issues will still exist in juno... side effect of people not understanding mongo and well... mongo
15:43:36 <cdent> And in general I think the project as a whole has a bit of an image problem that is not helped by the fact we don't often think about what we're doing in operational terms. Instead funcional.
15:43:52 <eglynn> I'd like to let the thread play out a bit more, then respond acknowledging the issues in the storage layer and point to the efforts to improve scalability & perf with gnocchi
15:44:25 <eglynn> anyone who'd like to rebut very specific points also (e.g. the mysql by default), obviously feel free
15:45:06 <fabiog> eglynn: we should ask for deployment scenarios where these people found issues
15:45:28 <fabiog> eglynn: those could be real pointers to where we need to be in the future regarding performane
15:45:39 <cdent> I really think that actively engaging with operators to figure out how to improve ceilo would be a great thing. Real use cases, etc.
15:45:41 <eglynn> fabiog: in terms of deployment sizes etc.?
15:45:41 <jd__> I think we are all well aware of our issues and we're trying to address them, we just lack manpower
15:45:48 <cdent> I'm not sure how we ...
15:45:51 <cdent> yeah what jd said
15:46:09 <fabiog> eglynn: exactly. So we will know what they expect and at which scale
15:46:15 <cdent> However, unless we are actively, almost constantly, telling people "we're working on it" no one will think we actually are
15:46:15 <jd__> we have ~10 developers with an equivalent of 2-3 developers full time max I'd say
15:46:25 <jd__> cdent: sure
15:46:50 <jd__> fingers crossed the situation is going to be far better once Gnocchi enters the scene
15:46:50 <eglynn> true, but there's only so many times you can say "patches are welcome" and not get much response
15:47:11 * cdent nods
15:47:42 <cdent> some small part of that is probably people rocking up and saying "i'd like to help but architecturally this is a mess so nevermind"
15:48:05 <cdent> (that is, of the very small number of people who interested, some segment is turned away in distaste)
15:48:13 <cdent> (gnocchi ought to help that)
15:48:31 <gordc> cdent: architectually what is a mess?
15:48:35 <eglynn> yes, distaste for archiecture ... aslo with some social aspects IIRC
15:48:48 <cdent> gordc: ceilometer
15:49:24 <cdent> So yeah: image problem. Which is fixable, but takes some work, of what sort I'm not sure.
15:49:37 <cdent> Work that is especially hard to do given resource constraints.
15:49:58 <gordc> cdent: uh.. ok. i'm not sure anyone would know what to do with that if someone just said that and dropped the mic.
15:50:34 <cdent> Well I'm not trying to finish the topic, I'm trying to start it.
15:50:55 <cdent> But on the other hand I don't think we can finish it right here right now. We just need to put it in our heads and care.
15:51:08 <eglynn> we need to demonstrate progress with gnocchi integration, provide realistic hope for a "game changer" IMO
15:51:23 <eglynn> it doesn't have to be absolutely the finished article in kilo
15:51:34 <eglynn> but enough to realistically tech preview
15:51:35 <gordc> cdent: yeah, i'm just saying i don't know what i would do with that if you said it to me... it's way way too broad.
15:51:43 <jd__> we need to polish ceilo/gnocchi integration and let people play with it
15:51:47 <cdent> which "it" gordc?
15:51:52 <eglynn> jd__: exactly, I agree
15:52:00 <cdent> jd__++
15:52:06 <gordc> cdent: "architectually, this is a mess"
15:52:07 <jd__> so gogogo pick item up in the k3 list and work on it! :p
15:52:24 <eglynn> ^^^ wot that man said!
15:52:40 <cdent> gordc: I'm saying that I think people have that reaction. The reasons why they do are unclear.
15:52:54 <linuxhermit> I can say that I've heard that a few times
15:52:58 <cdent> Surely it doesn't surprise you that they might have that reaction?
15:53:24 <gordc> cdent: not saying they're wrong. i'm just saying it'd help if they gave the reason so we could fix it.
15:53:33 <jd__> I understand that the architecture seems like a mess because it grew organically in different direction with a large scope
15:53:44 <jd__> thing is we are improving that
15:53:59 <cdent> We're talking about two different things. The issue is not that the architecture is a mess.
15:54:13 <jd__> I think we expose our plan a lot, e.g. we write about it and we do webinar (hint hint eglynn) but people just don't listen :)
15:54:14 <eglynn> cdent: you're saying the issue is perception?
15:54:16 <jd__> they want _results_
15:54:19 <cdent> yes eglynn
15:54:31 <cdent> I think the architecture is a mess, but I think we can fix it.
15:55:00 <eglynn> cdent: the way to change that perception is to show concrete progress on a realistic game changer, IMHO
15:55:08 <cdent> But an operator/user cares about whether they can percive that there is progress and a good direction.
15:55:12 <cdent> And I agree with you eglynn
15:55:17 <linuxhermit> jd__ as a new operator I've not seen those things
15:55:27 <cdent> exactly linuxhermit
15:55:38 <cdent> there is _zero_ visibility into the real future of ceilo
15:55:56 <cdent> (for people who aren't tracking patches)
15:56:32 <eglynn> cdent: I gave an RDO hangout earlier in the week and put gnocchi absolutely front and centre
15:56:48 <cdent> great, good start
15:56:49 <eglynn> cdent: ... i.e. I said nothing else about all the other work that's been landing in kilo
15:57:04 <fabiog> cdent: also there isn't an articulated and phased description on how Gnocchi and Ceilometer will merge/interact. Maybe we can start work on this
15:57:07 <rbowen> Video at https://plus.google.com/events/cht3k5nr5u73pv3d08i7vq5m570
15:57:19 <cdent> fabiog: that's a good idea too
15:57:26 <eglynn> (not becaus ethat wasn't important, more coz we need to continually bang the gnocchi drum)
15:57:43 <eglynn> ... in order to start the perception shift
15:57:53 <cdent> I'd prefer to state what I'm trying to say in a more encouraging way but I get the impression I'm making people defensive, that's not my intent at all.
15:57:55 <linuxhermit> nod
15:58:31 <cdent> We don't need to start Perl 6.
15:58:33 <eglynn> cdent: sorry if that came across as defensive ... that was not intended on my part, and I'm glad you brought this up
15:59:41 <fabiog> cdent: yes, thanks. It really helps to have ops perspectives. At the end of the day  *they* are our customers :-)
16:00:02 <eglynn> so we need semi-closure on gnocchi and we need to get the message out on it at summit
16:00:16 <gordc> so i guess track the thread and hope they give specific issues that need to be fixed?
16:00:18 <cdent> with regard to making gnocchi visible and popular I do think we can do a better job of having the "articulated and phased description" that fabiog mentions.
16:00:20 <eglynn> semi-closure == not necessarily done & dusted, but enough to play with
16:00:54 <cdent> that will also help drive contributions appropriately
16:01:24 <cdent> times up
16:01:26 <fabiog> eglynn: I will be happy to work on it. I took responsiblity for documenting this anyway from the list
16:01:40 <eglynn> fabiog: excellent, that would be welcome
16:01:45 <eglynn> yep, we're hogging the channel beyond our allocated slot
16:01:56 <eglynn> thanks folks and apols for the late start
16:01:58 <gordc> laters
16:02:00 <fabiog> eglynn: I will need to meet with you and jd__ to discuss it in detail
16:02:06 <linuxhermit> bye all
16:02:08 <eglynn> #endmeeting ceilometer