15:00:54 #startmeeting ceilometer 15:00:56 Meeting started Thu Aug 14 15:00:54 2014 UTC and is due to finish in 60 minutes. The chair is eglynn. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:57 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:58 o/ 15:01:00 The meeting name has been set to 'ceilometer' 15:01:03 o/ 15:01:03 o/ 15:01:07 o/ 15:01:08 o/ 15:01:16 o/ 15:01:31 <_nadya_> o/ 15:01:36 hi y'all! 15:01:36 o/ 15:01:40 #topic Juno-3 planning 15:01:44 o/ 15:01:54 can we do a quick round-table on the high/medium proirity BP status? 15:02:00 #link https://launchpad.net/ceilometer/+milestone/juno-3 15:02:18 ildikov and I discussed discussed dispatcher-for-gnocchi-integration earlier 15:02:28 conclusion was to bump this off the juno-3 slate because 15:02:42 .. dispatcher code will land in stackforge/gnocchi as opposed to the openstack/ceilo repo 15:03:01 ... also gnocchi is still decoupled from the main release cycle, so juno-3 is not a drop-dead date 15:03:23 ildikov_: that pretty much sums it up, amiright? 15:03:40 eglynn: yeap, I think you told everything 15:03:57 ildikov_: cool, thanks! 15:04:09 gordc: how's it going with bigger-data-sql? 15:04:12 eglynn: /me thanks for the quick summary 15:05:20 cdent: have you got any interesting detail on bigger-data-sql? 15:05:57 Nothing huge to add, other than that discussion we had about the tests maybe being too biased in the mongo direction, resulting in some SQL that is painful 15:06:26 We'll want to decide whether we should do anything about that. 15:06:41 on the multi-source/user/project aspect per resource? 15:08:01 k, no need to go into that now I guess 15:08:13 nsaje: central-agent-partitioning is looking pretty good at this stage, amiright? 15:08:20 eglynn: yep 15:08:26 nsaje++ 15:08:28 cool :) 15:08:30 me likely that code 15:08:31 eglynn, yes, that's much progress 15:08:31 everyone: needs code review :-) 15:08:34 likey 15:08:37 nsaje, np :) 15:08:48 llu: all good with snmp-improvement? 15:09:05 llu: (other than the need for review attention I guess) 15:09:48 ok llu might not be on, we can follow up on that tmrw morning 15:09:58 nealph: what about paas-event-format-for-ceilometer? 15:10:14 eglynn: work is progressing. Other than sphinx/tox install issues, there are no blockers. 15:10:19 On that...my sense is that everyone has similar issues with building docs, yes? 15:10:53 nealph: what issues are you having? 15:10:58 nealph: I can try to help you 15:11:05 nealph: excellent :) 15:11:11 ildikov_: thanks! 15:11:18 dependencies...thanks ldikov_! 15:11:32 prad: around? 15:11:37 nealph: I spent some time with fixing some Sphinx stuff earlier, let's see if I still remember those things :) 15:11:37 nealph, please share your problems :) 15:11:41 eglynn, hello 15:11:43 nealph, we can help you 15:11:51 prad: lots of discussion on the meter names for instance-per-disk-measurement ... headed for consenus, do you think? 15:11:52 DinaBelova: :) 15:12:34 eglynn, yes, i think we're close.. addressed most concerns in the latest patch.. once we come to agreement on resource naming in the docs we should be good to go i think 15:12:47 prad: excellent, thanks! :) 15:13:12 cdent: grenade-resource-survivability seem to have gotten over the ceilo service start-order hump in devstack, nicely back on track 15:13:37 <_nadya_> eglynn, prad, no more questions from my side about this patch (naming in docs doesn't seem too hard to manage with :) ) 15:13:40 That's right. With that devstack fix in place I should be able to tweak up the javelin2 changes, in line with the refactoring you suggested, and get that put to bed. 15:14:24 cdent: cool, have the QA folks given any indication of javelin being "re-opened for new features"? 15:14:51 there's fairly continuous work on it, with various tiny changes being merged all the time 15:15:04 so I'd give tentative "yes" on that, or at least "rsn" 15:15:30 cdent: cool 15:15:59 for taskflow-alarm-partitioning, I think we're gonna rebase this on the nsaje's new tooz-based parititioning scheme 15:16:24 blissful ignorance++ 15:16:27 nsaje, btw, how're you testing your tooz-based central agent HA ? 15:16:50 I mean, did you try some scenarios on how to test its workability 15:16:52 ? 15:17:21 DinaBelova: not yet, currently it's still theoretical :) 15:17:28 nsaje, a-ha, okay 15:17:30 DinaBelova: but if tooz holds its end of the bargain, it should work 15:17:44 idegtiarov is currently working on preparing the env to test it 15:17:49 nsaje ^^ 15:17:50 DinaBelova: that's amazing 15:18:05 DinaBelova: yeah, I didn't have time to prepare a test env yet, busy coding ;) 15:18:11 nsaje: the difficulty with testing is the lack of ZK packages? 15:18:18 because actually we need zookeeper in HA + multiple central agents + some scenarios, etc. 15:18:35 like we kill this process, what happens next and now to ensure it actually works 15:18:47 DinaBelova: yep, agreed 15:18:48 eglynn: yes, but we can also use memcached 15:19:05 nsaje: cool, that would be a useful confirmation of the approach 15:19:10 nsaje, we can, but ZK seems to be more preferable solution imho 15:19:43 but anyway :) 15:20:01 yep, from what jd__ said at mid-cycle the tooz ZK driver is definitely prefered for production 15:20:22 DinaBelova: perhaps, but we're using tooz in quite a light-weight fashion so I don't think memcached would present a problem (this is pure speculation, mind) 15:20:47 nsaje, yeah, probably, I just try to think what I'd love to see in prodaction possibly 15:20:52 production* 15:21:00 that's just some thoughts about that :) 15:21:08 DinaBelova: if you guys get a test env up, you could test memcached vs zookeeper as well :) 15:21:35 nsaje, yeah, we are in progress, let's see what can we reach this week and early next one 15:21:46 yeah, that would be a useful comparison point 15:22:06 anhyhoo, shall we move onto Tempest? (not that we're talking about testing) 15:22:14 #topic Tempest status 15:22:21 :) 15:22:41 DinaBelova: I saw your os-infra/config patch on the mongodb job, cool :) 15:23:00 eglynn, yeah :) 15:23:21 I guess we may try it 15:23:33 #link https://review.openstack.org/#/c/114208/ 15:24:03 for others: that's change to make MongoDB job not experimental but usual non-voting check 15:24:39 so that'll allow us in future to have both SQL and MOngo tested in the gate for all the changes 15:24:51 as for the Tempest tests itself 15:24:59 we're blocked really https://review.openstack.org/#/q/owner:%22Vadim+Rovachev+%253Cvrovachev%2540mirantis.com%253E%22+project:openstack/tempest+status:open,n,z 15:25:03 one useful thing to test might be unskipping the nova-notification test and seeing if it passed regularly against the mongo job 15:25:20 eglynn, oh, yeah, that's a good point 15:25:28 i.e. to see if the notification-test issues can be pinned down to sql-a perf 15:25:33 because if so, we'll unblock in some way other changes 15:25:41 okay, let me do that :) 15:25:43 and try :) 15:25:47 DinaBelova: cool, thanks! 15:25:52 anything else on Tempest? 15:26:03 eglynn, I guess no actually 15:26:09 #topic TSDaaS/gnocchi status 15:26:33 jd__: anything to report from your side? 15:27:19 there was some activity on that ML thread 15:27:22 #link http://lists.openstack.org/pipermail/openstack-dev/2014-August/042080.html 15:27:35 not much, I've just started to work on the archive policy 15:27:48 and I'm writing an article about Gnocchi in general also to have something for people to read about and understand 15:28:09 jd__: thanks, that blog post will be really useful I think to spread the word :) 15:28:35 are you planning to include more details about why starting something new was better than the existing options? that's been a topic of interest lately, and I'd like to get in front of it before there are issues 15:28:38 jd__, that's really nice, thanks :) 15:28:42 dhellmann: yes 15:28:47 great 15:28:52 BTW I gave the pitch to the TC on gnocchi on Tuesday last 15:28:55 #link http://eavesdrop.openstack.org/meetings/tc/2014/tc.2014-08-12-20.02.log.html#l-408 15:29:20 ^^^ that included some background questions on why whisper/carbon was considered unsuitable 15:29:50 ... that jd__'s article will address 15:29:52 cool 15:29:54 and why does ceilo team is reinventing time-series DB :) 15:30:04 DinaBelova: ... and that too :) 15:30:06 :D 15:30:32 ... a serious outbreak of NIH syndrome has afflicted us all ;) 15:30:57 anything else on gnocchi, or shall we move on? 15:31:29 #topic Future of the integrated release ML thread 15:31:58 k, so there's a lot of discussion about disparate "how to fix the future" topics on that thread 15:32:09 * cdent had internet fail 15:32:18 but some of it has focused in on the integration status of ceilometer 15:32:27 e.g. http://lists.openstack.org/pipermail/openstack-dev/2014-August/042920.html 15:32:35 and http://lists.openstack.org/pipermail/openstack-dev/2014-August/042962.html 15:33:07 a-ha, I see 15:33:41 obviously I'm presenting the case for ceilometer, and will continue to fight that corner 15:33:56 eglynn, thank you sir 15:34:15 the discussion will continue, I believe the TC are planning a couple of extra meetings over the coming days 15:34:19 I reckon fighting that aspect is a bit of a red herring. 15:34:43 cdent: no, it's quite serious 15:34:45 cdent: which aspect is a red herring, d'ya think? 15:35:36 Well, apparently I'm wrong, but I would have thought that it would be _very_ hard to de-graduate something with demonstrated numerous deployments. 15:35:52 cdent, not so hard as you think I guess 15:36:02 And that that data ought to stand on its own merits and talking about it just feeds the trolls. 15:36:04 cdent: a vocal minority of the TC does not currently believe ceilometer should remain integrated 15:36:19 if we have deployment info, we need to make sure that's public 15:36:24 * dhellmann hasn't read the ML yet this morning 15:36:30 It ought at least to be a different thread. 15:36:45 And it would be good if the TC (and the vocal minority) was a bit more transparent :) 15:36:48 cdent: part of the problem is that the argument has involved shifting of the goalposts (... and may well continue to do so) 15:36:57 cdent: this is an evolving conversation 15:37:16 dhellmann: I referenced the production deployment numbers from the Juno user survey on the ML 15:37:23 eglynn: great, that'll help 15:37:56 the fact that there are several stackforge projects that overlap with ceilometer doesn't help, unfortunately 15:38:33 dhellmann: do you mean Monasca? 15:38:34 Maybe, then, it is important to distinguish between ceilometer and telemetry. 15:38:55 Because ceilometer is an implementation which _may_ be lame, but telemetry is a clear need. 15:39:05 eglynn: and stacktach and there's another I don't remember off the top of my head 15:39:11 And I, at least, think of myself as working on telemetry. 15:39:39 cdent: the argument these people are making is that because ceilometer isn't good enough, another team should be given a chance to prove they are 15:39:49 Synaps? https://wiki.openstack.org/wiki/Synaps 15:39:50 I don't agree with that at all, I'm just relaying what I'm hearing. 15:39:50 dhellmann: that's probably Monasca you're thinking of, used to be called HP Jahmon 15:40:05 ah, sorry, that's not on stackforge 15:40:19 nsaje: Synaps did come up in one conversation I've had, but yeah. Also healthnmon, but I don't know where that lives any more. 15:40:24 eglynn: that's the most recent, yes 15:40:31 dhellmann: Even if that were come to pass, wouldn't we all just meander over to whatever that new project is and, by the logic, ruin it too? 15:40:49 anyway, I don't want to get everyone down, I just want to make sure everyone understands this is not an idle discussion 15:41:25 dhellmann: yep, understood, and thanks for the TC insight on that 15:41:26 cdent: I'm not sure they're thinking that far ahead 15:41:53 People are strange. 15:42:14 I guess I just wanted to bring this up on everyone's radar, so that we're all aware of it 15:42:36 I guess it's about numbers now for the TC, I mean deployments, etc :( 15:42:47 ... but by the same token we should try to not allow it become a distraction/demotivation factor for juno-3 15:43:04 Has anyone tried the "Hey, why can't we all just get along, come join the party" tactic? 15:43:19 several times 15:43:22 replacing one project with another is a very lame way to waste resources 15:43:23 bummer 15:43:25 ildikov_: yes, so hard data on non-small production deployments would be useful 15:43:33 right - the best thing to do is keep working on fixes and testing to bring things up to par 15:43:34 cdent, I guess we all share your opinion here 15:43:45 cdent: but everybody tries to convince the other guys to pick their stuff and stop their pet project 15:44:02 * cdent is not wed to anything 15:44:13 and also I guess bringing Gnocchi in is not really helpful in this situation 15:44:57 ildikov_, btw, fair point - like doing smth new before fixing something old one.... 15:44:57 ildikov_: well, the fact that we've a plan to address architectural debt should be a positive IMO 15:45:15 eglynn, yes, that should be shown in nice way 15:45:17 i.e. we're not sticking our head in the sand and ignoring the issues 15:45:17 ildikov_: the new API is an improvement, but there are questions about the wisdom of building gnocchi, which is why jd__'s blog post will be helpful 15:45:30 eglynn, the message here should be clear 15:45:44 dhellmann, ++ 15:45:46 the TC should note I'm absolutely not blocking them from solving the problem in any other way 15:45:56 jd__: "them"? 15:46:03 eglynn: from that point yes, it's just still not a proven change 15:46:11 s/them/people from the TC/ 15:46:35 dhellmann: +1, any post that makes this change and the reason for this change is really useful now 15:46:36 or anybody FWIW 15:47:10 <_nadya_> unfortunately agree with ildikov_. But I hope we are wrong :) 15:47:38 _nadya_: yeap, I hope the same :) 15:47:46 jd__: so the fact that gnocchi is being built out at arm's length from ceilo should be seen by the TC as a net positive IMO 15:48:35 eglynn: +1 - the fact that ceilometer will support other time-series databases is good, assuming we can get them running in the CI infrastructure for functional testing 15:48:50 _nadya_: the unfortunate in the current situation that people like seeing the dark side first, before thinking anything good about something or at least that's my experience, that's why I mentioned this PoV 15:48:50 I'd just like to point out that it's easy to build a specialized product with pick-of-the-litter backends, not quite so easy to build it on abstractions allowing a plethora of different backend combinations 15:49:41 nsaje: point being? 15:49:48 dhellmann: yeah, I made that point clear to the influxdb guy at the mid-cycle, i.e. the importance of getting the packages/RPMs in the ubuntu and EPEL6 repos 15:49:57 nsaje: that's a great point. we need to make sure the history is clear, because not everyone has followed the discussions we've had with other projects about sharing or integrating or replacing parts of ceilometer 15:50:03 jd__: that whoever is building alternatives to Ceilometer has it easy 15:50:15 eglynn: good, that will be key, as we've seen with mongodb in the past 15:51:23 ok, I guess we're mostly on the same page with this, I guess we'll have to see how the TC discussions play out this evening 15:51:41 nsaje: got it :) 15:52:05 anyone got anything else they want to raise up on this? 15:52:12 that meeting is in just over 4 hours if anyone else wants to listen in (it's not just about ceilometer at this point) 15:52:26 sorry, 3 hours 15:52:47 dhellmann: I'll definitely lurk on that 15:52:52 good 15:53:15 dhellmann: on this channel? 15:53:33 ah, no, #openstack-meeting-3 15:53:43 dhellmann: ok, thanks 15:54:00 dhellmann, ok 15:54:09 k, let's move on so, up against the shotclock 15:54:25 #topic godc to take on PTL reins from Aug 18-28 15:54:37 * eglynn will be on PTO from Aug 18-28 15:54:46 eglynn, good for you :) 15:54:53 :) 15:54:57 DinaBelova: :) 15:54:58 a well deserved break :-) 15:55:05 dhellmann, ++ 15:55:10 not great timing I'm afraid, but anyway ... 15:55:18 gordc has kindly agreed to take up the PTL duties 15:55:37 While you are gone we are going to conspire to destroy ceilometer. 15:55:52 cdent: LOL :) 15:55:53 eglynn: there's never a good timing, so please don't work during those 10 days :) 15:55:54 <_nadya_> hehe 15:55:56 cdent, let's be silent :) 15:55:59 :) 15:56:27 so anyhoo, please keep gordc in the loop on BP status etc. if anythng goes into the weeds 15:56:39 #topic Open discussion 15:56:39 eglynn, ok, gotcha 15:57:05 3 mins left ... 15:57:26 * cdent moves to adjourn 15:57:55 grapefruit 15:57:55 * eglynn seconds the motion 15:57:59 2... 15:58:16 I guess that's it :) 15:58:18 k, thanks for all for your time! 15:58:20 #endmeeting