15:01:17 #startmeeting ceilometer 15:01:18 Meeting started Thu Jun 12 15:01:17 2014 UTC and is due to finish in 60 minutes. The chair is eglynn. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:19 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:21 The meeting name has been set to 'ceilometer' 15:01:27 o/ 15:01:28 o/ 15:01:29 o/ 15:01:30 o/ (50%) 15:01:35 o/ 15:01:41 sorry for the slightly late start ... 15:01:44 o/ 15:01:44 o/ 15:01:48 o/ 15:02:07 #topic juno-2 planning 15:02:13 o/ 15:02:22 first things first, juno-1 is out! :) 15:02:32 #link https://launchpad.net/ceilometer/+milestone/juno-1 15:02:42 thanks for your efforts folks! :) 15:03:09 most of what we'd target'd for juno-1 actually made it, which is good :) 15:03:32 hbase-resource-rowkey-enhancement bumped to j2, but should land once the gremlins in the gate are squashed 15:04:02 ... also had to bump good patches for LP #1326250 #1326723 #1293337 to j2, due to gate lag :( 15:04:03 Launchpad bug 1326250 in ceilometer "swift broken when ceilometer middleware is enabled" [High,In progress] https://launchpad.net/bugs/1326250 15:04:27 btw - we actually merged first pack of tempest tests this morning - as tempest is branchless and etc. - we have our first tests in j1, really :) 15:04:47 DinaBelova: excellent! 15:04:56 eglynn, ;) 15:04:58 the dream come true :) 15:05:20 so we are on our way :) 15:05:36 DinaBelova: \o/ :) 15:06:09 so, other than the bumpees from juno-1 ... 15:06:20 the cupboard is currently looking slightly bare WRT juno-2 ... 15:06:23 #link https://launchpad.net/ceilometer/+milestone/juno-2 15:07:09 a very rough laundry list of stuff that may be in the game for j2 follows ... 15:07:20 eglynn: I have a proposal that could land there - https://review.openstack.org/#/c/95817/ 15:08:12 fabiog: thanks, I'll look off-line 15:08:21 eglynn: IIRC, there were some separate BP comments in the Gnocchi spec too 15:08:48 ildikov: got a link? 15:08:53 eglynn: I'm not sure if any of those could fit into j-2 though 15:09:32 eglynn: I haven't checked if they were even created after, one of these is the pipeline design 15:09:50 ildikov: not sure what you mean by "BP comments in the Gnocchi spec"? 15:10:30 eglynn: that it will be specified in a separate BP, like how the pipeline design will look like 15:10:48 a-ha, ok 15:10:54 eglynn: but maybe we just simply not there yet, I will check later 15:11:08 so here's a very rough list of possibilities ... 15:11:19 sql-a viability phase2 15:11:25 documentation improvements 15:11:32 event persisence in mongo 15:11:42 code-free meters 15:11:48 SNMP improvements 15:11:55 central agent repartition 15:11:56 eglynn: what do you mean by docco improvements? 15:12:14 eglynn: can we discuss central agent repartition a bit more? 15:12:34 ildikov: basically the TC gap coverage task 15:12:50 Alexei_987: one sec, lemme just run thru' the rough list quickly 15:12:57 consume IPMI sensor data from ironic 15:12:58 k 15:13:03 network services metering 15:13:10 grenade phase2 (javelin) 15:13:16 per-project pipeline 15:13:23 complex queries for stats 15:13:27 eglynn: it's already a BP in OS Manuals and agreed to fit in j-2 somehow, so it's not on our list 15:13:35 gnocchi/ceilo-core integration 15:13:47 ildikov: k 15:14:19 eglynn: complex query waits for Gnocchi until someone tells me that it should be implemented for the current design 15:14:23 tempest testing - also j2, as I remember 15:14:38 eglynn: the pipeline design for Gnocchi could fit into the last topic you mentioned 15:14:51 DinaBelova: ... yep, tho' does it need a BP also? 15:14:53 ildikov: yep 15:15:03 eglynn: and on top of that can come the dynamic config and the per-project pipeline 15:15:04 eglynn - tempest spec 15:15:05 :) 15:15:13 so probably no in the ceilo 15:15:43 so basically, if any you intend to work on any of the above item and there isn't already a BP spec 15:15:49 ... please propose one to ceilometer-specs 15:16:05 did I leave out anything? 15:16:26 hi, I'd like to implement the bp https://blueprints.launchpad.net/ceilometer/+spec/advanced-policy-rule in j-2, not sure if there is any conflict to our plan 15:16:42 which is not approved yet :( 15:16:50 zqfan: have you proposed a spec to ceilometer-specs? 15:16:50 eglynn: I think it's a long enough rough list already ;) 15:17:06 not yet, that is registered long time ago 15:17:19 so in general, here's how ttx wants the BP process to work now ... 15:17:27 1. propose spec to ceilo-specs 15:17:34 2. core team review & approve 15:17:54 3. use my awesome spec2bp to turn the spec into a blueprint 15:17:59 3. ceilo-drivers or PTL files a corresponding BP on launchpad and links to the spec 15:18:09 a-ha, ttx, even better! 15:18:13 ttx, heh - magic :) 15:18:16 (not ready yet) 15:18:23 eglynn, ok, will upload the spec tomorrow 15:18:29 that was a teaser, sorry for interrupting :) 15:18:40 zqfan: ... so the point is, the entry point is now ceilo-specs, not the launchpad BP 15:18:42 ttx: where can we learn that magic? 15:18:56 llu-laptop - teaser :( 15:19:03 eglynn, ok 15:19:08 as ttx said :) 15:19:13 llu: will be in openstack-infra/release-tools when complete 15:19:25 :( 15:19:28 further ttx magic includes an auto-kick script that'll untarget any BPs that don't follow the above workflow 15:19:47 Alexei_987: so you wanted to discuss central agent repartition? 15:19:57 eglynn: small POC here https://github.com/temujin/ceilometer-parallel-agent/commit/cd154f22a0b15c7bf2f6dcbbe72404c71ca206bf 15:20:04 eglynn: can you please explain a little bit more about "code-free metering"? 15:20:20 eglynn: slight change in processing flow that allows us to fork central agent without external HA tools 15:20:33 KurtRao: discussed at summit https://etherpad.openstack.org/p/ceilometer-snmp-inspector 15:20:39 eglynn: so no need in master election 15:20:47 Alexei_987 - so you just make it stateless? 15:20:52 Alexei_987: any reason you want to avoid using taskflow? 15:20:58 KurtRao: (second half of that etherpad) 15:21:07 gordc: additional complexity in configuration 15:21:25 gordc: it's much easier to handle it on our side 15:21:30 gordc, I guess Alexei_987 is trying to avoid tooz :-\ 15:21:51 no the taskflow 15:21:58 DinaBelova: any external lib brings configuration, maintanance, compatibility burden 15:22:13 if we can avoid using it why not? 15:22:28 eglynn: got it. Very interesting feature 15:22:34 Alexei_987: so the plan was to use taskflow/tooz for central agent & partitioned alarm eval 15:23:05 eglynn: And my opinion is that we can change our algorithms so it won't be needed 15:23:28 Alexei_987: how? 15:23:34 just like POC provided 15:23:44 Alexei_987: can you propose your PoC to gerrit? ... and we can continue the discussion there? 15:23:47 llu-laptop: using stuff that we already have available 15:23:54 eglynn: sure we can 15:24:26 eglynn: need to update tests so 15:24:39 eglynn: my main question is do we want it or not? 15:24:57 hi all, would you please approve this https://review.openstack.org/#/c/95526/ 15:25:01 that's based on a database to distribute the tasks? 15:25:07 jd__: yes 15:25:10 Alexei_987: i'd hold off on building tests 15:25:48 gordc: why? 15:25:54 ok, do not want 15:26:00 jd__: why? 15:26:00 Alexei_987: i think we all need to take a look at it first... so you don't waste your time 15:26:14 gordc: you can look on github in such case 15:26:19 and leave comments there 15:26:25 koteswar16: that's a neutron patch? 15:26:30 gordc: on gerrit it will not pass jenkins 15:26:58 Alexei_987: k, if it's proposable to gerrit, let's do a pre-review on github 15:27:11 ... it's not proposable to gerrit 15:27:33 move on? 15:27:40 ok 15:27:42 Alexei_987: yeah... but just saying. people will have questions about design alone... tests might be jumping the gun since plans might change.... if you're free i guess you can code tests. 15:28:01 eglynn: works for me... we can comment on design offline. 15:28:06 #topic Gremlins in the gate 15:28:10 so a major spike in our py26 unit test failure rate 15:28:10 eglynn: +1 15:28:17 http://jogo.github.io/gate/ 15:28:30 due to https://bugs.launchpad.net/ceilometer/+bug/1328679 15:28:31 Launchpad bug 1328679 in ceilometer "string io errors in unit test" [Medium,Triaged] 15:28:45 sileht: white smoke on https://review.openstack.org/99654 ? 15:28:56 eglynn, yes 15:29:05 sileht: \o/ :) 15:29:20 eglynn, because of eventlet we should never have timeout in tests 15:29:43 sileht: ... so can you un-WIP the patch? 15:30:08 eglynn, I want to do a tone of recheck to be sure it sufficient 15:30:14 eglynn: am i reading graph right? the failures have disappeared as of today? 15:30:20 eglynn, I can't reproduce it locally 15:30:22 sileht: cool, wise 15:30:36 gordc: scroll down to the 3rd graph 15:30:42 sileht: why we can't have timeout, it's a eventlet semaphore 15:30:58 gordc: sorry that's py27 ... my bad 15:31:00 eglynn: that's py27 15:31:02 @Sylvain Afchain: would u please approve this https://review.openstack.org/#/c/99654/ ?? 15:31:07 eglynn: or just because we're not running tests... so it tapered off. 15:32:08 llu-laptop, in case of test_notification is the first test to run, then eventlet pause it (because oslo.messaging use time.sleep), then all others tests are run (that take more than 2*30 seconds) 15:33:05 llu-laptop, this is my guessing 15:33:47 sileht: so it sounds plausible, I guess we just need a few more green rechecks to give further confidence? 15:34:06 eglynn, yes 15:34:47 eglynn, but I think this is unreallated with the py27 gate issue, because this one seems fail only in py26 15:35:03 sileht: yeah I've only seen it on py26 15:35:12 sileht: cool enough... so once you're confident, let's try to get that puppy landed 15:35:37 move on? 15:35:56 #topic TSDaaS/gnocchi update 15:36:12 jd__: the floor is yours, sir! 15:36:22 thanks 15:36:35 So as promised I've started to integrate with Ceilometer 15:36:44 cool 15:36:55 the PoC has been developed as a dispatcher driver 15:37:08 so it plugs at the collector level, and tries to map the samples received to what Gnocchi expects 15:37:10 https://review.openstack.org/#/c/98798/ 15:37:21 *cough*, *cough* ... is that somthing that could be written up in a BP spec possibly? 15:37:28 it's early stage and for now it just create the instances resources in Gnocchi that's it 15:37:34 eglynn: probably yeah 15:37:43 eglynn, ts-s! don't tell jd__ about the specs :) 15:37:46 DinaBelova did a nice schema to represent that for people interested 15:37:52 * jd__ stares at DinaBelova for the URL 15:38:02 * DinaBelova looking for it 15:38:17 https://docs.google.com/a/mirantis.com/drawings/d/1OMqtrgtvWLu_SYEh-g8K0iiIC6lrBWhn8G6KGGIkpS4/edit?usp=drive_web 15:38:17 #link https://docs.google.com/drawings/d/1OMqtrgtvWLu_SYEh-g8K0iiIC6lrBWhn8G6KGGIkpS4/editDinaBelova: 15:38:22 :D 15:38:24 I'm not sure I'll make a lot of progress in the next days as I've other things on my plate 15:39:05 other than that, I'll do a hangout to do a walkabout of the source code for people interested in jumping in 15:39:06 jd__: do you wanna mention that code walk-thru' tenatively planned for monday? 15:39:11 snap! 15:39:25 #link https://plus.google.com/events/c7nbeqejlp0ot3lh3trss11meok 15:39:29 @eglynn: Yes neutron patch. but related to ceilometer labels... 15:39:40 so feel free to join 15:40:05 thanks jd__! 15:40:12 that's all boss 15:40:14 jd__, cool, thanks 15:40:39 #topic Ceilometer sub-sprint at QA/Infra mid-cycle meetup in Darmstadt, Germany 15:40:51 jd__: thanks, this walk-thru is a really good idea! 15:41:04 I mentioned this QA mid-cycle on the channel during the week 15:41:15 ... bringing up here again for completeness 15:41:30 mtreinish floated the idea of using one of the QA/Infra mid-cycle days as a ceilometer testing sub-sprint 15:41:45 #link https://wiki.openstack.org/wiki/Qa_Infra_Meetup_2014 15:41:50 (prolly July 17th or 18th) 15:41:53 eglynn: do we have a deadline to decide? 15:42:05 not a hard deadline 15:42:17 ildikov - the thing is we need at least two ceilo ATCs to be there... 15:42:42 exactly ... critical mass >= 2 ceilo folks 15:42:43 I hope to attend the Paris event - so I'll have no chance to be in Darmstadt 15:42:49 if so 15:42:50 :) 15:43:07 so just wanted to throw out the idea on the table so folks can think about it 15:43:14 DinaBelova: well, not impossibly big number at least :) 15:43:25 ildikov, yes, for sure :) 15:43:27 eglynn: I may can convince my managers 15:43:43 cool 15:43:50 ildikov - if so, please tell me 15:43:55 eglynn: I wanted to catch up with testing anyway, it seems to be an effective way 15:43:56 gimme a shout over the next couple days if anyone else would be interested 15:44:04 I'd like to discuss this tempest topic before you'll be there :) 15:44:07 ... it's a bit tentative still 15:44:52 move on? 15:45:02 @eglynn: can some one approve my change? 15:45:05 #topic Open discussion 15:45:10 DinaBelova: sure, I will use the cat look from Shrek, if nothing else works ;) 15:45:16 ;) 15:45:19 koteswar16: we're in a meeting, can it wait? 15:45:31 ok 15:45:38 koteswar16: check back on the Ceilo channel after the meeting 15:47:20 ... i guess we're going to finish on time for once? :) 15:47:31 gordc: LOL :) ... looks like it 15:47:33 gordc: LOL :) 15:47:40 sweet! 15:48:31 ... and the world cup doesn't even start for another what, another 4 hours? :) 15:48:45 gordc: we deserve it after the overbooked last 3 or more meetings :) 15:49:14 eglynn: pub chosen or beer already in the fridge at home? ;) 15:49:32 ok that's a wrap ... thanks for a productive meeting 15:49:39 #endmeeting ceilometer