15:00:11 #startmeeting ceilometer 15:00:12 Meeting started Thu Oct 2 15:00:11 2014 UTC and is due to finish in 60 minutes. The chair is eglynn. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:13 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:15 The meeting name has been set to 'ceilometer' 15:00:20 o/ 15:00:29 hey, y'all 15:00:48 o/ 15:00:54 o/ 15:01:48 #topic Juno close-out 15:01:56 o/ 15:02:00 so RC1 is in the bag! :) 15:02:06 o/ 15:02:12 o/ 15:02:21 ... or tagged and bagged as the cool kids would say :) 15:02:23 #link https://launchpad.net/ceilometer/+milestone/juno-rc1 15:02:53 so the final juno tag will be based on this 15:03:04 *if* we don't find any showstoppers in the mean time 15:03:34 <_nadya_> o/ 15:04:04 if we do need an RC2, then the fix would have to be landed on master first then backported to proposed/juno branch 15:04:08 eglynn, I just wanted to point one interesting bug 15:04:15 DinaBelova: shoot 15:04:19 possibly it'll be a huge pain if we won't fix it 15:04:24 * DinaBelova searching 15:04:52 eglynn, https://bugs.launchpad.net/python-ceilometerclient/+bug/1357343 15:04:53 Launchpad bug 1357343 in python-ceilometerclient "ceilometer-alarm-evaluator fails after sometime with giving 401" [Medium,In progress] 15:05:26 it's the issue having roots in the our client updated to the oslo common code 15:05:44 so now after long time alarm evaluator running 15:06:03 DinaBelova: token expiry? 15:06:07 eglynn, yeah 15:06:08 :) 15:06:38 meh! we had similar problems before with the ceiloclient, which were fixed like a year ago 15:06:50 OK, that definitely needs to be fixed 15:06:50 eglynn, yeah, so there was the change by me merged to oslo-incubator 15:07:02 and now it's trying to be updated in the ceilo client 15:07:05 one moment 15:07:11 DinaBelova: so we will need an RC2 in that case 15:07:30 eglynn, here it is https://review.openstack.org/#/c/125058/ 15:08:01 although it's kind of blocked for now with gordc change 15:08:19 eglynn: are clients under the same release rules? 15:08:40 eglynn, it won't be needed the rc2 15:08:50 it'll be ceilo client change 15:08:55 DinaBelova: my change merged... but i think we need to properly resync that item. 15:08:56 but we need to merge this asap 15:09:03 and create new ceilo client release 15:09:05 a-ha, I see 15:09:07 for the installations 15:09:39 gordc, DinaBelova: the clients were supposed to be frozen before juno-rc1, but agreed, we'll definitely need a 1.0.12 to get this fix 15:09:47 eglynn, a-ha, ok 15:09:55 gordc, good to know it'll be merged 15:09:58 eglynn, could we please consider this for juno rc https://bugs.launchpad.net/ceilometer/+bug/1374012 15:09:59 Launchpad bug 1374012 in ceilometer "Ceilometer polls lbaas resources even when the neutron enabled and disabled lbaas" [Medium,Fix committed] 15:10:55 prad: yes, we should get that in 15:11:03 ty sir 15:11:13 OK looks like we will need both a new ceiloclient release and an RC2 15:11:29 eglynn, yeah, with this bug - yes 15:12:20 there's two other bugs i've tagged with rc-potential: https://bugs.launchpad.net/ceilometer/+bugs?field.tag=juno-rc-potential 15:12:51 https://bugs.launchpad.net/ceilometer/+bug/1369124 was the one that we thought was fixed in RC1 right? 15:12:54 Launchpad bug 1369124 in ceilometer "syslog in gate-swift-dsvm-functional full of ceilometer errors" [Medium,In progress] 15:13:22 eglynn: yeah, that was the one i where i fixed nothing... or i fixed half of it. 15:13:34 yeah, let's try to get our value in cutting an RC2 and get all 3 fixes in 15:14:09 eglynn: agreed. https://bugs.launchpad.net/ceilometer/+bug/1375568 actually only affects juno (it's a py26 related bug) 15:14:13 Launchpad bug 1375568 in ceilometer "Python2.6: Error importing module ceilometer.ipmi.platform.intel_node_manager: 'module' object has no attribute 'OrderedDict'" [Medium,In progress] 15:14:54 let's aim for EoD Monday for all fixes landed if possible, so that I can request the RC2 in the rel mgr 1:1 on Tuesday at 11:45UTC 15:16:14 two patches: https://review.openstack.org/#/c/124686/ https://review.openstack.org/#/c/124916/ 15:16:37 gordc, a-ha, ok 15:16:50 one last thing about the juno closeout 15:17:08 I need to report back to the TC on the gap analysis outcome 15:17:35 #link https://wiki.openstack.org/wiki/Governance/TechnicalCommittee/Ceilometer_Gap_Coverage 15:18:13 I've marked pretty much everything as completed 15:18:28 ... except one task under Tempest: "curate the ceilometer Tempest coverage to ensure it does not become persistently problematic in the gate" 15:18:44 ... which I guess could be interpreted as an ongoing requirement in any case 15:18:51 eglynn: seems accurate to me. 15:18:59 eglynn, yes 15:19:08 eglynn: the #4 in the docco is not done yet 15:19:26 there's incremental progress with getting the USE_SCREEN=False stuff turned on which is supposed to help with the problem with services sometimes not starting 15:19:29 eglynn: but I don't think that currently it would hurt anyone though 15:19:46 ildikov: a-ha, k, I'll try to "finesse" that one :) 15:19:57 there was a huge mess trying to get that stuff backported to icehouse (you may have seen Sean ranty email) but it has made it through just today 15:20:07 the backport was required to get grenade working 15:20:26 in the end this ought to result in more stable ceilo in tempest 15:20:29 eglynn: thanks, well, it's still better to have info on more places, than nowhere :) 15:21:00 cdent: nice work! I didn't see that ranty mail, but I'll read after the meeting, thanks for the heads-up 15:21:15 the gist is that icehouse in the gate is fuxored 15:21:21 and nobody cares 15:21:26 cdent, heh.. 15:21:57 DinaBelova: I wondering also about "land the stalled Tempest patches" task given that the nova notification test was then skipped? 15:22:08 DinaBelova: ... your unskip patch didn't land yet, amiright? 15:22:16 eglynn, that's true.. 15:22:38 DinaBelova: have you tried rebasing/running it again recently? 15:22:39 eglynn, here it is https://review.openstack.org/#/c/115212/ 15:22:54 gordc, will do today 15:23:00 had a hot weeks 15:23:12 DinaBelova: cool cool, np 15:23:24 cool, it would great to get that landed 15:23:30 eglynn, yeah, for sure 15:24:08 anything else juno related? 15:24:45 #topic TSDaaS/gnocchi status 15:25:03 jd__: the floor is yours! 15:25:30 hey 15:25:39 good progress again this week 15:25:50 we now have merged all archive policy stuff 15:26:00 cool :) 15:26:05 yay! 15:26:11 we have Keystone middleware enabled by default 15:26:18 (and support loading more middleware if needed) 15:26:33 sileht progressed on the Ceilometer dispatcher 15:26:35 * eglynn will finally rebase his influx patch to try to emulate the archive_policy logic 15:26:38 and started working on aggregation 15:26:55 and we fixed a bunch of bugs and race condition in the tests 15:27:00 so it's getting pretty solid at this stage 15:27:13 jd__, thx for the resume :) 15:27:30 yeah, interesting discussion on the mean-of-means versus mean-of-raw-data on the cross-entity aggregation review 15:27:38 we also now gate on py34-postgresql 15:27:50 (we can't gate on py34-mysql because mysql does not work with Python 3) 15:28:14 a-ha, so can that job be skipped? 15:29:07 ok, we only have a py34-postgres right now 15:29:17 so no need to skip, cool, got it 15:29:25 yup 15:29:51 also I'm working on the tooz IPC issues that sileht discovered 15:30:43 jd__, this one is not easy 15:31:27 BTW amalagon is working with the new archive policy support in her gnocchi custom aggregators 15:31:39 (... so that she can select the most granular data available when aggregating across periods, to avoid the mean-of-means distortion where possible) 15:31:48 sileht: finger crossed I'll fix it :D 15:32:00 jd__, I guess we can fix it, if gnocchi run as standalone (even with multiple worker), but for guys that will use wsgi, it's not possible 15:32:16 sileht: I'm actually trying to fix in tooz directly 15:32:41 does that issue only manifest with the posix_ipc driver for tooz? 15:32:52 eglynn, yep 15:33:17 would we be recommending the tooz/ZK driver instead for production deployments? 15:33:37 once you have more than one gnocchi node you don't have the choice 15:33:58 (or tooz/memcache driver) 15:34:33 yeah for now IPC is only for one instance of Gnocchi running or in the unit tests 15:34:34 so really IPC approach is mainly intended for really small deployments and the tests? 15:34:40 but I hope to fix that in tooz directly 15:34:46 eglynn: yes, one node or tests 15:34:54 cool, got it, thanks! 15:35:02 anything else on gnocchi? 15:35:23 #topic Tempest status 15:35:49 main thing here would be to get that nova notification test unskipped before the TC review if poss 15:36:23 (slated for the next TC meeting, Oct 7th) 15:36:43 eglynn, I'll rebase it 15:36:47 let's give it a try 15:36:48 :P) 15:36:57 DinaBelova: great, thanks! 15:37:02 eglynn, np 15:37:36 #topic kilo summit planning 15:37:58 yay, kilo sumit soon :) 15:38:03 so I got confirmation that we'll definitely be down from 10 to 6 slots for Paris 15:38:18 :( 15:38:23 that's akward... 15:38:30 but anyway :) 15:38:31 apparently the average cut was 33% as there's one less day of formal design sessions 15:38:49 eglynn, and that organised 'pod' day, yeah? 15:38:50 but that was weighted by a rough metric of "project activity" over juno 15:39:06 i.e. a metric of BPs/bugs/reviews etc. 15:39:12 so we got a slightly larger cut 15:39:27 I guess becuase some of the focus was on gnocchi 15:39:27 :( 15:39:33 eglynn, wow! i did not know they were using these stats 15:39:51 DinaBelova: me neither, until I asked how the decision was made 15:39:58 eglynn, a-ha, ok, got it 15:40:34 the upside is we'll have a full day on Friday of the "contributor meetup" ... i.e. the pod++ 15:40:38 that's going to be yet anothe perverse incentive 15:40:58 "I get more goodies if I make lots of useless bps/bugs/reviews" 15:40:59 cdent: yeap, as if we didn't already have enough of those ... 15:41:17 eglynn, I also wantd to note that ityaptin is preparing the lab to test gnocchi (speaking about the performance part, but who knows, probably we'll find some other issues) 15:41:29 enough of those perverse incentive I meant ... as opposed to useless BPs ;) 15:41:39 DinaBelova: coolness :) 15:41:43 :) 15:43:21 we'll discuss concrete topics next week 15:43:40 (as agreed last week) 15:44:09 moving on ... 15:44:10 I cannot attend the next week's meeting :( 15:44:11 #topic OPW ideas 15:44:46 ildikov: ok, we can maybe push it out a week, I'll check when the schedule needs to be formalized 15:45:11 eglynn: cool, thanks 15:45:24 on the OPW, we've a good record of diversity promotion on this project :) 15:45:49 two previous OPW interns (Terri & Ana) and another volunteer interested for the next round 15:45:52 eglynn, here is the etherpad https://etherpad.openstack.org/p/ceilometer-opw-ideas 15:46:07 so let's fill it :) 15:46:15 DinaBelova: thanks! ... exactly 15:46:28 I hope to spend time tomorrow for this thing 15:46:34 :) 15:47:02 if anyone has any project ideas, no matter how wacky, please do drop a quick description onto that etherpad 15:47:50 to give a sense of the scoping ... Terri worked on adding group_by semantics to the ceilo v2 API, Ana is working on period-spanning stats for gnocchi 15:48:45 thanks to DinaBelova for stepping up to mentor in this upcoming cycle! 15:48:54 eglynn, np :) 15:48:57 I'd love to try :) 15:49:05 I hope I'll be useful in this role 15:49:42 I've one for Gnocchi I think 15:49:55 jd__: nice one! 15:50:12 jd__, it should be telemetry :) gnocchi or ceilo - it does not matter imho 15:50:16 some people expressed interest in having a Ceph driver for Gnocchi 15:50:33 jd__, wow :) who are they? :) 15:50:37 just interested 15:50:38 :) 15:51:02 direct native ceph, as opposed to ceph-sitting-behind-swift? 15:51:06 eglynn: yes 15:51:20 cool, that sounds really interesting 15:51:21 jd__, that's interesting thing btw 15:51:25 eglynn, indeed 15:51:36 I talked about that with nijaba_ TBH 15:51:49 jd__, a-ha, ok 15:52:09 jd__: would it be an appropriate level of complexity/challenge for an intern? 15:52:20 eglynn: good question, I don't know 15:52:32 * nijaba_ confirms 15:52:34 eglynn, speaking about the technical thing - yeah, it would be 15:52:43 but it won't be clear OpenStack task to be honest 15:52:48 I think it is technically 15:53:11 cool, esp. if the intern has a bit of storage-foo already 15:53:13 yeah it's kind of a mix, though it's still writing code for OpenStack 15:53:18 like it would be writing a driver for Nova or Neutron 15:53:37 and since it would be based on Carbonara, it wouldn't be uber complicated too 15:54:08 looks closer than the etherpad after the first quick read 15:54:21 ildikov, yeah, for sure 15:54:21 cool, so the main challenge would be mapping the swift client semantics onto the equivalent under the ceph API? 15:54:28 eglynn: likely 15:54:32 ildikov, I just hope to find even better variant 15:54:46 is there a deadline for deciding the task? 15:55:36 Given recent hullabaloo about "vendory" stuff is a ceph project ideal? 15:55:37 ildikov: application are due by Oct 22 15:56:34 DinaBelova: well, it's never an easy task, I guess we could investigate a bit the Ceilo-Gnocchi integration area, maybe it could give some more tasks, I'm not sure now 15:56:41 cdent: ceph being open-source mitigates the vendory feel 15:56:43 ? 15:56:52 eglynn: cool, thanks 15:57:09 ildikov: but that the application deadline for the intern herself 15:57:45 eglynn: and they choose project or task? 15:58:00 eglynn: the difference between the two is one using rest api and the native rados library is not (it's a library calling directly the ceph nodes) 15:58:00 ildikov: yeah ... to be realistic, the idea would have to be firmed up before then as the intern applies for a particular project idea 15:58:09 eglynn: that's certainly an argument to make, I suppose, but some people would probably not agree and what's the win of just ceph v swift over ceph? 15:58:10 eglynn: but i don't think this is hard 15:58:19 eglynn: hmm, then we're already late :) 15:58:20 chmouel: a-ha, got it, thanks! 15:59:08 ildikov: yeah, sooner the better re. the ideas 15:59:13 I don't have a position on this, just observing. 15:59:32 eglynn: yeap, sure, got it 15:59:33 looks like this ceph work has definite potential though 16:00:10 time has run out 16:00:18 anything else on letting a thousand diverse flowers bloom in the open source world? :) 16:00:46 k, let's skip open discussion this week, the shot-clock has beaten us 16:00:54 thanks for your time folks! 16:01:02 #endmeeting ceilometer