21:02:56 #startmeeting ceilometer 21:02:57 Meeting started Wed Oct 23 21:02:56 2013 UTC and is due to finish in 60 minutes. The chair is eglynn. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:02:57 That's a sad time. 21:02:58 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:03:00 The meeting name has been set to 'ceilometer' 21:03:05 jd__ is on vacation so I'll fight with the ircbot this week ... 21:03:12 (here's hoping no-one calls for a vote on anything ... ;)) 21:03:15 o/ 21:03:17 o/ 21:03:19 hello 21:03:20 o/ 21:03:20 o/ 21:03:20 You will win where we previously failed. 21:03:23 o/ 21:03:29 o/ 21:03:45 any more hands to go up? 21:04:06 right-o, let's dig into the agenda items ... 21:04:15 #topic Ceilometer errors in tempest runs 21:04:26 #link https://wiki.openstack.org/wiki/Meetings/MeteringAgenda 21:04:26 or "when can we stop using sqlalchemy as the default backend in devstack" ;) 21:04:40 dhellmann the floor is your's ... 21:04:46 I thought sileht worked out what to do about that already 21:04:58 yeah we discussed this a while back on IRC with sileht & jd__ IIRC 21:05:05 ok 21:05:06 can you summarize? 21:05:08 we should be able to switch back to mongodb as the default in devstack 21:05:17 (as mongodb-server 2.4.x has hit the ubuntu cloud archive) 21:05:21 ftr, it looks like what we need to do is increase some of the defaults for pool sizes 21:05:23 * eglynn digs for the log ... 21:05:24 * dhellmann looks for the review link 21:05:36 #link http://eavesdrop.openstack.org/irclogs/%23openstack-metering/%23openstack-metering.2013-10-02.log 21:05:41 #link https://review.openstack.org/#/c/53396/ 21:06:00 dhellmann: should we be increasing them in ceilometer code? 21:06:16 gordc: we want ceilometer's defaults to be useful out of the box 21:06:22 dhellmann: well that may solve one issue with using sqlalchemy, but there are others 21:06:33 dhellmann: for example the lack of support for metaquery 21:06:38 eglynn: ok, this was the one I knew about :-) 21:06:48 that should just result in skipped tests, right? 21:06:51 (needed by horizon and ceilo alarming, possibly other too ...) 21:06:53 i'm kind of torn on that personally... because the defaults we set may or maynot work based on how many meters the environment is capturing. 21:07:16 well /me thinks we should be defaulting to our most feature complete storage backend 21:07:18 gordc: sure. but the tempest environment is very very small, so our standard defaults are clearly not enough 21:07:27 and the one that distros are likely to recommend for production 21:07:43 eglynn: sure. we need to figure out how to get the right version of mongodb into devstack then 21:07:52 dhellmann: yep, exactly 21:08:05 dhellmann: i guess... what's setting the defaults currently? the sqlalchemy code from oslo? 21:08:07 Wouldn't that be mongodb-10gen? I'm not sure how ubiquitous that is. 21:08:18 dhellmann: so according to sileht, mongodb-server 2.4.0 is now in the ubuntu cloud archive 21:08:23 (as per IRC link above) 21:08:33 eglynn: ok, does that mean the devstack gates will install it? 21:08:41 previously IIRC the prob was that it was only availble for precise from the 10-gen repo 21:08:41 gordc: yes 21:08:51 dhellmann: yep, it did up to now 21:08:53 eglynn: yes, that's right 21:09:04 dhellmann: sorry, up to being disabled as the default 21:09:07 ok, well, I think we should go for both then 21:09:26 Oh, I see. Yeah, like I run on Debian and the .deb for mongodb on Squeeze was <2.2. 21:09:40 So, I have to use mongodb-10gen when I swap to that backend. 21:09:50 right, developer environments are OK because we can write instructions for upgrading 21:09:56 yep 21:09:56 yep, so I'll follow up with sileht on that again tmrw on how to enable grabbing from the cloud archive 21:10:03 but the automated environment has different rules, so we need to make sure we're allowed to require the package 21:10:14 eglynn: sounds good 21:10:36 the infra folks didn't want us to require a package from the 10-gen repo IIRC 21:10:40 do we also want to accept a patch to raise the defaults? 21:10:46 eglynn: yes, that's right 21:10:48 but now we should be to switch back 21:10:50 cloud archive may be ok 21:11:03 be ok to switch back, yep exactly 21:11:14 eglynn: we should probably do some work to figure out how out of sync the drivers are w.r.t. features 21:11:23 because I think the mongo driver does not deal with events, right? 21:11:32 dhellmann: yep agreed 21:11:33 maybe I missed that being added, though 21:11:43 dhellmann, there are reviews up for it 21:11:50 dhellmann: i'm ok with raising defaults.... only thing is the tests i ran, we can/have way more than 10 sessions running at any given time. 21:11:53 let me snag them 21:11:53 would be good to go through before the summit 21:11:58 there's a patch in the review queue also to add metaquery to sqlalchemy 21:12:08 #link https://review.openstack.org/#/c/49548/ 21:12:19 #link https://review.openstack.org/#/c/49456/ 21:12:23 https://review.openstack.org/#/c/49456/ 21:12:28 snap! :) 21:12:30 ;) 21:12:42 gordc: we can work out the actual values through gerrit discussion; I don't know the right answer there 21:12:54 dhellmann: cool cool. 21:13:16 so this will bite us again I think unless we become firmer re. feature parity between the drivers 21:13:35 eglynn: yeah, I think that should be a goal for icehouse 21:13:37 eglynn, +1 21:14:01 and where we don't have parity, we need a very clear statement of that 21:14:01 we should do something to specify which parts of the driver API are related, maybe with separate base classes or something to hold the interfaces? 21:14:22 i.e. the implications of chosing a storage backend in terms of what stops working 21:14:31 dhellmann: yep, that would make sense 21:14:50 eglynn: we could also look at some of the interface description libraries other than abc 21:14:59 Yeah. The tests can be a good source for that. If your driver can't pass this series of tests, it's out of compliance. Sort of like that timestamp ordering one. 21:15:13 but I feel like we have a lot of "this method is implemented, but not all aspects of it work" kinds of differences 21:15:14 for resources 21:15:22 especially with variations on queries 21:16:07 so we might want to think about whether it makes sense to split those out into separate methods, for clarity 21:16:24 I hate even suggesting that, though :-/ 21:16:50 dhellmann, I'm not sure I follow. 21:17:04 nova has separate methods like "find resource by id" and "find resource by name" 21:17:07 yeah, its a non-trivial question, how best to capture the functional requirements imposed on the storage driver by various features 21:17:10 we have "find resource" with optional parameters 21:17:20 and then in some drivers one of the parameters doesn't work, so we raise NotImplemented 21:17:52 yep, suddenly pagination or metaquery blows up when you switch drivers 21:17:59 that's a simplistic example, the real example is more like we don't support a particular way of querying against the metadata associated with an event in some cases but not in others 21:18:06 right 21:18:06 I see. Yeah, that's difficult to say which would be deserving of that or not. Some are just simply filters, others may have completely different query structures. 21:18:38 we don't want to say we can't add features until all the drivers work, but we should probably say that at least one driver must always have all of the features 21:18:48 and pick a single driver to be the main or default or whatever 21:18:51 In any case, you could get a bit clever with your tests and see that the method behaves the same across drivers. 21:18:54 dhellmann: +1 21:19:00 that used to be mongo, until we had a bunch of people not wanting to use it add features :-) 21:19:04 * dhellmann looks at sandywalsh 21:19:15 Lol, he's on a plane right now, I'd imagine. 21:19:31 +1 for maintaining mongo as the canonical driver 21:19:31 thomasem: we do, in fact, run the same test code against all of the drivers 21:19:47 dhellmann, I know. What I meant was flesh out in the tests what a compliant driver looks like. 21:19:58 I would go along with that, or even with sqlalchemy, but either way we should pick one 21:19:59 with what* 21:20:23 thomasem: ok, I'm not clear on what you're suggesting we would do different 21:20:36 drivers are allowed to raise NotImplementedError for missing features, and the tests just skip 21:20:37 dhellmann, Not different, just ore of. :) 21:20:45 ah, ok, yes to more testing :-) 21:20:48 dhellmann, Take that timestamp ordering test for example. 21:21:10 how about we try to grab an unconference slot for this discussion at summit? 21:21:14 that would provide a reference point...but would we then require those adding new drivers to update mongo with any missing functionality? 21:21:14 (to involve the RAX folks with skin in the game re. events) 21:21:23 +1 on the session 21:21:43 * eglynn is assuming events is the main incomplete aspect of mongo currently 21:21:44 eglynn: not a proper conference session in the ceilometer room? are we out of slots? 21:21:54 eglynn: did we figure out how many slots we have at summit? 21:22:07 gordc: 11 last I heard 21:22:09 * thomasem is becoming sadder by the second that he can't join y'all. 21:22:10 nealph: yes, the idea would be all new features have to land in mongo if they are added to another storage driver 21:22:19 or, if not mongo, the driver we do agree on 21:22:35 dhellmann: +1 21:22:41 dhellmann: me& sandywalsh will be there @ hk. I'm up for an unconf session. 21:22:52 dragondm: cool 21:23:08 should we move on? 21:23:19 eglynn: why an unconference session, instead of a regular session? 21:23:34 dhellmann: I'd be up for either 21:23:39 15 proposals so far....some trimming already needed. 21:23:45 I'll propose one on this 21:23:52 dhellmann: 21:23:55 dhellmann: cool! 21:23:59 #topic summit update 21:24:00 nealph: there's a few regarding api so i would think they could be merged. 21:24:25 eglynn: dhellmann:, I'd be up for either. I'm sure sandy would as well. 21:24:25 on the subject of HK, the ceilo track for the design sessions will be finalized shortly by jd__ 21:24:37 too late to propose another? 21:24:41 we're over-subscribed, so I suspect there will be some merged sessions 21:24:59 (some obvious candidates for merging among the proposals) 21:25:06 #link http://summit.openstack.org/cfp/details/355 21:25:42 dhellmann: thanks! 21:25:53 so jd__ will be finalizing the approved sessions and pushing the schedule to sched.org shortly 21:26:09 (this time round, IIUC the design summit and conference will have completely separate schedules) 21:26:20 and apparently run in different buildings too ... 21:26:38 I think it's separate floors of the same building, but a very very large building 21:26:47 there is quite a distance between the two spaces, I think 21:26:51 a-ha, ok 21:26:59 (apparently some of the other projects had problems before with randomers showing up at design sessions) 21:27:03 that wasn't entirely clear from ttx yesterday, though, so I could have that wrong 21:27:12 * eglynn doesn't remember it being an issue with ceilo though 21:27:13 that should be interesting...I've seen only a couple of cm-specific conference talks though. 21:27:14 yeah, the nova room in particular was packed 21:27:49 what constitutes a 'randomer'? non-contributer? 21:27:59 it's a shame I'm going to have to miss so many of the ceilometer sessions to be in the oslo room 21:28:06 makes me wonder if the design summit & conference shouldn't be 2 separate events altogether... 21:28:15 gordc: yeah, we had non-atc folks clogging up the doorways 21:28:21 gordc: good question, maybe a business-y type in a suit? ... I dunno 21:28:26 dragondm: "bringing developers and users together"... 21:28:38 dhellmann: gotcha. 21:28:53 heh, ok I guess there's a point there :-> 21:28:54 eglynn: i'll remember to lose the suit ;) 21:29:07 anyway last I heard our track runs from late Tuesday afternoon, thru' all day Wed, into Thursday morning 21:29:31 (no doubt y'all are planning your schedules for the week ...) 21:29:31 eglynn: yeah, that's what it looks like on the latest layout form I've been given 21:29:52 dhellmann: yep, partially clashing with oslo 21:30:18 I think we overlap for 5-6 sessions 21:30:23 Oy. 21:30:30 yeah, I won't be there for much of wednesday 21:30:31 dhellmann: so if that "Feature parity in storage drivers" session is approved 21:30:50 dhellmann: (we'd want it to be shceduled for a non-overlap-with-oslo slot) 21:30:53 yeah, jd__ will have to juggle a bit 21:31:00 yeap 21:31:02 as I've done for the sessions he proposed for oslo :-) 21:31:41 it looks like he has 6 slots to work with for that, so shouldn't be an issue 21:31:48 cool, intra-PTL horse trading ;) 21:32:25 :-) 21:32:35 just a quick reminder also that jd__, nijaba_ & myself are doing a conference session on ceilo+heat=alarming at 14.20 on Friday 21:32:38 * dragondm is now imagining jd_- juggling horses... 21:32:49 #link http://openstacksummitnovember2013.sched.org/event/02a1ea504b3dec538eaedca5e31461ed 21:32:59 How amusing! 21:33:01 ... we got the graveyard shift ;) 21:33:09 eglynn: was just going to ask for link 21:33:34 BTW do we want to pre-arrange a ceilo evening meet-up? 21:33:48 I guess we could just glom onto the eNovance event on the thursday evening ... 21:33:59 * eglynn is a freeloader ;) 21:34:19 #link http://openstacksummitnovember2013.sched.org/event/6d7b3e0f9778fa675cd7f0ea56becba3 21:34:49 if memory serves we talked about doing a social event at Portland 21:35:02 but everyone had too many other parties to go to ... 21:35:25 we could plan to meet at the enovance party :-) 21:35:36 * gordc is down for a party. 21:35:41 cool, sounds like a plan 21:35:42 oops, should have read the backlog :-) 21:36:04 * terriyu wonders what developer parties are like 21:36:36 Depends on th budget :_> 21:36:41 terriyu: lasers. 21:36:51 terriyu, unicorns 21:37:19 yep a lot depends on the budget 21:37:28 unicorns are expensive ;) 21:37:34 are you guys going to pull that trick where you tell me to wear certain clothes and I'm only one who shows up to the party dressed that way? 21:37:34 thomasem: that's all you need unicorn and lasers :)... never been to one myself so looking forward to it. 21:38:04 Just the horse mask. 21:38:04 haha 21:38:11 This is the truth. Because if you can actually get a unicorn, you've probably already had enough to drink. 21:38:20 we usually wear meter reader uniforms, right? 21:38:26 LOL 21:38:49 yep, with matching peaked caps, very fetching :) 21:38:52 I just imagined that. It was amazing. Will y'all do a synchronized dance too? 21:39:07 ooo, if not this time, definitely in the spring 21:39:21 * dhellmann wants to see jd__ leading us in a line dance on stage 21:39:28 * thomasem does too. 21:40:21 #action jd__ choreograph a line dance for summit 21:40:22 * dhellmann can't wait to hear jd__'s reaction after reading the meeting log from this week 21:40:26 haha 21:40:37 I'm sure it'll be grand. :) 21:40:49 sorry to lead everyone off topic, I guess this is what happens when jd__ is not here ;) 21:40:53 eglynn, did we have anything else on the agenda? 21:40:55 #topic any other business 21:41:14 or open discussion as its known around these parts ... 21:41:17 I saw something on the agenda about release python-ceilometerclient 21:41:23 I don't know if that is still relevant 21:41:27 that's a recurring topic 21:41:36 we should verify that there's no need for an update, though 21:41:38 ah yes 21:41:39 #topic release python-ceiloclient? 21:41:40 I don't think so, is there? 21:41:49 don't thinks it's required 21:41:55 seconded 21:41:57 the patch to update the global requirements to 1.0.6 finally landed 21:41:59 eglynn: something went in ... i don't remember what it was 21:42:13 so that 1.0.6 could be pulled in by heat & ceilo itself 21:42:21 (the latter for the alarm evaluator) 21:42:32 eglynn: nm unit test fix. 21:42:48 gordc: a-ha ok 21:43:06 so no need to re-release then for a test update 21:43:15 agreed 21:43:27 #topic open discussion 21:43:39 * eglynn has had fun this week kicking the tyres on the new horizon admin metering panel 21:43:49 props to lsmola for getting that landed for havana! :) 21:44:00 found a few nits but should all be nicely ironed out before 2013.2.1 21:44:10 (i.e. first release off stable/havana) 21:44:26 eglynn, cool 21:44:31 Oh, that's awesome. I'll have to go check that out. 21:44:41 also related to the UX ... 21:45:02 one of the women-in-openstack intership ideas for icehouse is sparklines capturing the recent trend for resources 21:45:06 is the new metering panel also in devstack? 21:45:20 #link https://wiki.openstack.org/wiki/OutreachProgramForWomen/Ideas#Sparklines_in_Horizon 21:45:24 cool 21:45:37 terriyu: it should be, if it's in horizon 21:45:40 terriyu: it's in havana, so defo on master 21:45:51 terriyu, yes I install it via devstack 21:45:52 terriyu: (note though, admin only for now) 21:46:14 terriyu: (not visible if you log in as a joe-soap user) 21:46:25 cool, thanks for the info guys 21:46:43 eglynn: cool intern project, thanks for pointing it out 21:47:29 terriyu: yeah, could be quite neat visual candy for new users 21:47:49 (possible on-ramp to more folks paying attention to ceilo ...) 21:47:51 eglynn: I hope one of the interns picks that project :) 21:48:01 terriyu: yep, me too! 21:48:15 it would be cool 21:48:29 not much else from me ... 21:48:40 oh, one other thing 21:48:49 lsmola: since you're a mentor, you can drop by #openstack-opw and say hi to the applicants 21:48:49 last 'normal' release off stable/grizzly ... https://launchpad.net/ceilometer/+milestone/2013.1.4 21:49:11 from now on, stable/grizzly will prolly be limited to security fixes 21:49:34 eglynn: did we resolve the havana wsme issue? 21:49:38 terriyu, I am already there, but i did't say hi :-) 21:49:41 please continue to tag havana bugs though for stable/havana if appropriate 21:49:58 terriyu, more like waiting for some questions :-) 21:50:13 lsmola: ok, I'll try to direct some applicants your way 21:50:15 gordc: we pinned the dependency version to < 5.5b6 IIRC 21:50:31 terriyu, ok, cool 21:50:50 eglynn: ah cool cool. i'll take a look. i just remembered we had a bunch of havana patches sitting around. 21:51:35 gordc: (that was on stable/grizzly that the wsme version is limited, on master we're up to latest I think) 21:52:18 eglynn: right... i guess i was mostly wondering about this: https://review.openstack.org/#/c/52426/ 21:52:35 gordc: here's the exact change https://review.openstack.org/#/c/52348/3/tools/pip-requires 21:53:10 gordc: ouch, didn;t realize that wasn't merged 21:53:29 eglynn: yeah... i don't think we ever resolved stable/havana 21:53:55 gordc: OK, thanks for the heads-up, I'll look into that tmrw 21:54:00 https://review.openstack.org/#/c/52602/ is just sitting there. 21:54:22 (apologies to everyone for bringing this up late... just remembered) 21:54:51 eglynn: cool, i'll sync up with you on that tomorrow... see what we've decided. 21:54:55 np! I'll try to upick that gordian knot tmrw ... 21:55:02 gordc: better late than never! 21:55:53 right-o, coming up near the end of our timeslot 21:56:04 anything else anyone wants to raise? 21:56:24 going once ... 21:56:35 going twice ... 21:56:49 #endmeeting ceilometer