15:00:10 #startmeeting ceilometer 15:00:11 Meeting started Thu Jan 29 15:00:10 2015 UTC and is due to finish in 60 minutes. The chair is eglynn. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:12 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:15 The meeting name has been set to 'ceilometer' 15:00:16 o/ 15:00:18 o/ 15:00:24 <_elena_> o/ 15:00:35 o/ 15:00:40 o/ 15:00:47 o/ 15:01:15 o/ 15:01:41 o/ 15:01:55 hey folks, I've been on semi-PTO all this week so only partially around 15:02:05 #topic Kilo-2 status 15:02:06 lucky you 15:02:17 LOL :) 15:02:22 #link https://launchpad.net/ceilometer/+milestone/kilo-2 15:02:44 so we've got a fair few BPs yet to land before next week 15:02:45 mmester: from the PTO PoV for sure :) 15:02:52 most are in pretty good shape 15:03:22 except the hyper-V related BPs are showing a great deal of progress 15:04:05 I'm planning to punt these to kilo-3 if we don't see something my EoW 15:04:11 *by EoW 15:04:41 any other concerns for kilo-2? 15:04:55 eglynn: the conf store bp will be addressed by 3 patches, 1 in kilo2 and 2 in kilo3 15:05:09 eglynn: the kilo 2 is ready for review https://review.openstack.org/#/c/142592/ 15:05:38 fabiog: so you've taken over shepharding from nealph_afk? 15:05:39 Is the base gabbi/http-test stuff enough to say phase one is done or should I lay in a few more tests? 15:06:01 eglynn: yes 15:06:31 cdent: the more tests the better, but what's proposed is prolly enough to declare vistory for k2 I think 15:07:09 fabiog: so I'll target this to kilo-2 in LP? ... https://blueprints.launchpad.net/ceilometer/+spec/ceilometer-configuration-via-data-store 15:07:23 I've moved on/over to gnocchi for the moment to see what issues a different evironment reveals. 15:07:32 eglynn: we are going to have 3 patches 15:07:34 cdent: coolness, makes sense 15:07:46 eglynn: the first is storage and we would like to land it in K2 15:08:05 then API and Agents and these will land in K3 15:08:12 makes sense? 15:08:23 eglynn, and API stuff will be implemented by idegtiarov for kilo-3 15:08:29 with client-side 15:08:47 fabiog: yep ... so when we discussed with this Phil before, IIRC we decided to have 2/3 blueprints in launchpad for tracking purposes 15:09:08 fabiog: the first BP was to capture what was planned for kilo-2 15:09:20 fabiog: the other one or two BPs were to capture what was planned for kilo-3 15:09:44 eglynn: ok, do I need to do something or these bp are already aligned? 15:09:47 fabiog: so is https://blueprints.launchpad.net/ceilometer/+spec/ceilometer-configuration-via-data-store currently the only LP BP registered for this? 15:09:59 eglynn: I think so 15:10:10 the patch partially implements it 15:10:16 eglynn, fabiog: https://blueprints.launchpad.net/ceilometer/+spec/configdb-api is registered for the api work. 15:10:37 nealph: was there a specific LP BP intended for the kilo-2 work? 15:11:18 eglynn: as you noted, the thought was to have the original bp targeted for Kilo-2 15:11:19 nealph: e.g. ceilometer-configuration-via-data-store==>kilo-2, configdb-api==>kilo-3? 15:11:28 that ^^^. :) 15:11:33 nealph: coolness 15:11:38 eglynn: can we use https://blueprints.launchpad.net/ceilometer/+spec/ceilometer-configuration-via-data-store and create another one for the agents? 15:12:22 fabiog: yep, I've just targeted ceilometer-configuration-via-data-store tp kilo-2 15:12:49 eglynn: ok, but there is the agent specific work that will go to Kilo-3 and it is dependent on the API 15:12:51 fabiog: ... and yes, please register a fresh one for the agents if it's a standalone tranch of work 15:12:57 eglynn: got it 15:13:01 thanks 15:13:25 fabiog: cool ... once you've registered the agents BP, I'll target to kilo-3 15:13:31 fabiog: thank you sir! 15:13:51 anything else for kilo-2? 15:15:01 cool enough, moving onwards ... 15:15:04 #topic gnocchi status 15:15:31 fabiog did an interesting g+ hangout earlier in the week on gnocchi measure batching 15:15:41 ... was that recorded BTW? 15:16:40 eglynn: not sure, I think it wasn't 15:16:49 but I posted the deck 15:16:52 jd__, eglynn - sadly I should mention I'm kind of stolen a little bit to one more Mirantis project for now, so I'll slow down work on gnocchi a little bit. I'll finish for sure my current patches, but I suppose that lots of my ideas will be going through ityaptin :) 15:17:19 I hope that's not more than for one month 15:17:21 fabiog: fair enough, yeah the slides will be good for anyone who missed to catch up 15:17:39 fabiog, I'm having going through them in my TODO 15:17:40 eglynn: http://www.slideshare.net/FabioGiannetti/gnocchi-batching 15:17:48 fabiog: Thanks! 15:17:49 the slides: http://www.slideshare.net/FabioGiannetti/gnocchi-batching 15:17:49 DinaBelova: fair enough, thanks for the heads up on that 15:17:57 fabiog, ildikov: thanks! 15:18:50 DinaBelova: I saw the mid-point reviews were due for the OPW, all good on that score? 15:18:56 fabiog - possibly you may take a look on one my gnocchi change (wip for now, hope to finish debugging to eow) - https://review.openstack.org/#/c/148271/ 15:19:14 DinaBelova: will do ;-) 15:19:35 eglynn, yes - and we've decided to focus on KairosDB driver 15:19:40 I have generic testing-related question that has come up as a result of prodding at gnocchi: to what extent is it important for the internal wsgi/pecan app to be robust in the face of missing data that can only be missing if keystonemiddleware is either not there or for some reason broken 15:19:43 that's enough module-maintaing thing 15:19:48 good research, etc. 15:19:54 for example if the x-user-id header has not been produced 15:20:02 I've spoken with opw folks, they are ok with this task 15:20:08 even if it was delayed 15:20:16 DinaBelova: coolness, thanks 15:21:59 cdent: I'm not sure I understand the question ... if x-user-id is not set, then the auth token verification would have failed, in which case the dispatch path should be cut short and rejected with a 403? 15:22:22 cdent: ... a-ha you mean if the keystone authtoken m/w is completely absent? 15:22:28 yes 15:22:43 because when testing the internal app you don't want to also be testing keystone 15:22:57 cdent, yeah... 15:22:58 DinaBelova: does kairosdb make justinb's cassandra work useless? 15:23:13 from metering pov 15:23:21 gordc, not fully, actually.... but - there is kind of intersection for sure 15:23:23 cdent: so for the first few months of its existence, the default mode for gnocchi was no keystone IIRC 15:23:45 cdent: ... and I don't remember that being problematic 15:23:51 gordc, anyway usage of kairosdb will need lots of efforts due to the same problems we're facing with opentsdb 15:24:00 (speaking about data retention, etc.) 15:24:00 the current concrete problem that I'm encountering eglynn, is that a sqlalchemy exception reaches the surface if you provide not x-user-id when creating a resource 15:24:20 DinaBelova: ok. i'll keep that in mind next time i speak with justinb 15:24:20 it seems that should be transformed to something less data layer exposing 15:24:38 but since it is "impossible" by some defintions, it may not be that relevant 15:24:52 gordc, yeah - so possibly in the neasrest future kairosdb driver will have the limited functionality, comparing with justinb directly-ceilo solution 15:25:02 that does not face pre-aggregation issues, etc. 15:25:31 DinaBelova: i see.. (i'll try to dig into it when i find some time) 15:25:38 cdent: a-ha, got it ... the created_by fields etc. were added *after* the keystone intregration was done, and that logic is probably not tolerant to the x-{user|project}-id headers being missing 15:25:51 gordc, yeah, feel free to ping me or ityaptin about this 15:26:25 DinaBelova: will do. 15:26:26 exactly eglynn so my question boils down to: do we regard that to be a bug 15:26:27 DinaBelova: kairosdb has no native downsampling support? 15:26:48 eglynn - it has the same big-table concept underlaying 15:27:02 DinaBelova: I'll may ping you too, as I'm just about the help eglynn to get the Influx driver in shape, so I thought to try to get more knowledge about OpenTSDB too, to see what are the common things 15:27:04 and all things like compaction are supposed to be done periodically 15:27:15 ildikov, yeah, for sure :) 15:27:47 eglynn ^^ and out of concrete storage layer 15:27:59 cdent: yes, I think it should be possible to work without keystone 15:28:02 cassandra and hbase have much in common in this case 15:28:18 eglynn, cdent - yep, otherwise it'll be the complete hell to test it 15:28:25 :( 15:28:27 DinaBelova: I was just wondering if there are some common behaviors of these DBs, then we might not want to rewrite them, but anyway, let's chat about that some time and then we will see :) 15:28:49 ildikov, yeah. let's do that 15:29:27 DinaBelova: cool, thanks in advance :) 15:29:36 ok, cool 15:30:08 ildikov: DinaBelova please include me in DB related discussions when you have one :-) 15:30:21 fabiog, ok, sure :) 15:30:35 fabiog: I planned to have it on the channel, will ping you then 15:30:57 ildikov: I can offer a phone bridge, if that helps ... 15:30:58 notable patches that landed this week include a new capability API to expose support aggregation methods and a simple carbonara CLI 15:31:01 https://review.openstack.org/#/q/status:merged+project:stackforge/gnocchi,n,z 15:31:02 fabiog: and of course we will schedule it to fit your TZ too :) 15:31:28 fabiog: the channel has the advantage of logging 15:31:42 and the devstack plugin is nice addition, anyone using that regularly? 15:31:43 ildikov: sure 15:32:19 fabiog: and I'm also in learning phase as for InfluxDB, so maybe we will not need voice at this stage, but thanks, it can be useful later, when we have some progress for sure 15:34:13 anything else on the gnocchi work? 15:35:14 k, moving onwards ... 15:35:33 #topic open discussion 15:36:41 who'd have thunk that Lemming-gate would blow up like that? 15:37:17 teapots 15:37:36 lemming-gate? 15:37:38 eglynn: one question: where new api features should be added in v2 or in v3 15:37:48 gordc: http://lists.openstack.org/pipermail/openstack-dev/2015-January/055338.html 15:37:55 eglynn: I was on a soft-skill traning in the last two days, I'm braindead ;) 15:38:15 ah one of those tl;dr emails 15:38:36 idegtiarov: well, the v2 API is going to be around for a while, so you can continue to improve it in incremental ways 15:38:52 eglynn: got it, thanks 15:38:57 ildikov: ... i.e. no need to consider it frozen as yet 15:39:12 gordc: yeah, drama-of-the-week 15:40:05 a quick note from me.... 15:40:14 gnocchi doesn't appear to have its own launchpad setup, so where does one report bugs? 15:40:21 eglynn: I meant that no thoughts should be expected from me, but reading it back it's really not that funny :) 15:40:37 nealph: the floor is your's sir! 15:40:56 just wanted to note that I'll be stepping away from ceilometer. :( 15:41:03 cdent, directly to jd__ )))) 15:41:22 nealph: :( 15:41:28 nealph: well thank you for your contributions up to now 15:41:42 certainly...have appreciated working with all you folks! 15:41:49 nealph, I was really happy to work with you on this project :) 15:41:50 nealph: Phil, thanks for all the hard work you put on the project. We really appreciate it! 15:42:04 who will be getting the benefit of your skills now? 15:42:30 cdent: still within HP, away from OpenStack I'm afraid. 15:42:31 nealph: It was cool to work with you on ceilo :) 15:43:05 I shall miss grousing with you when everyone else is at summit. 15:43:09 nealph: :( 15:43:18 cdent: ha, yes. 15:43:35 keep up the good work, all. :) 15:43:41 nealph: wish you luck and joy on the new area! 15:43:58 ildikov: thank you much! 15:45:17 cdent: a proper launchpad footprint will be required once jd__ has big-tented gnocchi under the openstack namespace 15:45:34 cdent: ... in fact, it should really have it now even under stackforge 15:45:52 (IIRC other stackforge projects such as tooz and packstack do) 15:47:26 anything else anyone? 15:48:05 is there any plans to get OVS stats into ceilometer? 15:49:07 mmester: I don't know of any specific plans, nothing on our immediate roadmap at least 15:49:35 mmester: how are these stats surfaced by ovs? 15:50:30 eglynn: not sure exactly. :/ 15:50:57 eglynn: ill work on it and put up a BP so at least we have a record of it somewhere 15:51:05 mmester: excellent, thank you sir! 15:51:43 k, let's close a few mins early so if there's nothing else 15:51:48 thanks folks! 15:51:54 #endmeeting ceilometer