14:01:05 #startmeeting watcher 14:01:06 Meeting started Wed Aug 10 14:01:05 2016 UTC and is due to finish in 60 minutes. The chair is acabot. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:01:08 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:01:10 The meeting name has been set to 'watcher' 14:01:11 hello ! 14:01:22 hi 14:01:24 o/ 14:01:28 hi 14:01:28 hi 14:01:34 agenda for today #link https://wiki.openstack.org/wiki/Watcher_Meeting_Agenda#08.2F10.2F2016 14:01:39 hi 14:01:48 o/ 14:02:00 #topic Announcements 14:02:08 a couple of announcements today ;) 14:02:25 #info Watcher release mode has been updated to cycle-with-milestones but TC prefer that we wait for the Ocata cycle 14:02:31 \o 14:02:35 #link https://review.openstack.org/#/c/353333/ 14:02:41 \o 14:02:42 hi 14:03:00 it doesn't change anything, we just have to do it again at the end of the newton cycle 14:03:23 #info ongoing discussions on the ML regarding Watcher use cases for Nova placement API 14:03:31 #link http://lists.openstack.org/pipermail/openstack-dev/2016-August/100600.html 14:03:38 #link http://lists.openstack.org/pipermail/openstack-dev/2016-July/100327.html 14:03:54 thx edleafe for describing our needs on the ML 14:04:19 jwcroppe : did you get feedback from jaypipes ? 14:04:34 acabot: yep, I updated the etherpad as well 14:04:49 #link https://etherpad.openstack.org/p/newton-watcher-midcycle 14:05:03 acabot: there is a lot of placement discussion going on - really two different ideas 14:05:37 jwcroppe : ok so anything we can do on our side for now ? 14:06:05 just follow the ML :-) 14:06:26 follow ML for now 14:06:30 ok 14:06:51 edleafe : any update on your side or the ML is up to date ? 14:06:53 yeah - nothing Watcher-related will be possible until the placement API is solid 14:07:13 edleafe : ok 14:07:37 edleafe : and when do you see that happened ? ocata ? p ? 14:07:40 there is no interest in adding a new API to live migrate if the placement API is coming soon 14:07:52 hmm 14:08:04 ocata is possible, but P is more likely 14:08:04 edleafe : yes it depends is the placement API is targeted for ocata 14:08:13 ok 14:08:14 hi 14:08:24 acabot: we're trying to get the basics in Newton 14:08:36 edleafe : ok 14:08:40 acabot: so that the data will be migrated by ocata 14:08:52 #info tpeoples stepped down from Watcher core 14:09:07 #info alexchadin is promoted as Watcher core, please vote on the ML 14:09:15 http://lists.openstack.org/pipermail/openstack-dev/2016-August/100828.html 14:09:18 #link http://lists.openstack.org/pipermail/openstack-dev/2016-August/100828.html 14:10:06 #info (special for edleafe ;-)) jellyfish is now the official Watcher mascot 14:10:26 ;-) 14:10:35 acabot: :) 14:10:38 #info votes for Barcelona talks are closed, we will get the results on August 25th 14:11:04 Jellyfish! nice 14:11:13 :) 14:11:15 #info last mid-cycle etherpad with actions #link https://etherpad.openstack.org/p/newton-watcher-midcycle 14:11:53 so please review your actions and strikeout them when completed 14:12:13 any other announcement ? 14:12:45 #topic Review Action Items 14:12:56 watcher-specs 14:13:20 alexchadin : you should update https://review.openstack.org/#/c/339456/ 14:13:51 acabot: I'm working on it and will discuss some things with jed56 14:13:55 alexchadin : I saw your quick chat with vincentfrancoise, is the scope clear enough ? 14:14:11 ok thx 14:14:26 just after the meeting ;-) 14:14:30 #action alexchadin update spec https://review.openstack.org/#/c/339456/ to reflect mid-cycle decisions 14:14:32 thanks! 14:14:51 https://review.openstack.org/#/c/298891/ needs reviews from core 14:15:05 because Greg's team is ready to release the code 14:15:20 #action jwcroppe sballe acabot review and merge https://review.openstack.org/#/c/298891/ 14:15:45 Added spec for automatic triggering audit needs a new PS 14:15:51 #link https://review.openstack.org/#/c/342473/ 14:16:17 I dont think digambar is here 14:16:43 #action digambar add a new PS for https://review.openstack.org/#/c/342473/ 14:16:54 Limit concurrent actions invoked by Watcher needs reviews 14:17:01 #link https://review.openstack.org/#/c/348049/ 14:17:09 jwcroppe : is this spec ready for reviews ? 14:17:50 watcher 14:17:58 Add goal_id, strategy_id and host_aggregate CLI options to audit has been merged 14:18:05 #link https://review.openstack.org/#/c/332982/ 14:18:10 acabot: yes 14:18:18 thx hvprash 14:18:25 acabot: it's ready for review 14:18:38 jwcroppe : ok thx 14:18:41 thx 14:19:09 Add Scoring Module implementation needs reviews 14:19:15 #link https://review.openstack.org/#/c/351577/ 14:19:33 I dont think tomasz is here 14:19:50 but he is probably waiting for reviews :-D 14:20:10 who wants to review the scoring module ? 14:20:14 I will 14:20:37 already started bu now tomasz added tests so I will go through them 14:20:39 i can do it also 14:20:51 #action vincentfrancoise jinquan-tmp review https://review.openstack.org/#/c/351577/ 14:20:54 thx 14:21:04 Added Monasca Helper needs reviews 14:21:10 hi, sorry for being late - have problems with network 14:21:11 #link https://review.openstack.org/#/c/348891/ 14:21:26 i can have a look 14:21:29 tkaczynski : no pb, is the scoring module ready for reviews ? 14:21:40 gzhai-mobl: thanks 14:21:41 yes, it's ready 14:21:50 gzhai-mobl : yes I'd like to have you on Monasca 14:22:20 #action gzhai-mobl review https://review.openstack.org/#/c/348891/ 14:22:43 tkaczynski : vincentfrancoise & jinquan will review your code, thx 14:22:59 Added strategy ID + Action Plan syncing needs reviews to close the efficacy-indicator BP 14:23:05 #link https://review.openstack.org/#/c/325213/ 14:23:20 hvprash : could you please take a look at this PS ? 14:23:36 sure 14:23:38 acabot: thanks. I'm not working this week ad the next one, but I will be addressing all potential issues to not block the review 14:24:07 tkaczynski : great thx, I'd like to have the scoring engine merged by August 28th 14:24:26 #action hvprash review https://review.openstack.org/#/c/325213/ 14:24:40 acabot: me too, I'll do my best to make you guys happy :) 14:24:52 :-) 14:24:54 :) 14:24:59 Add rally-jobs folder to get rally support any update since Rally additions ? 14:25:05 :p 14:25:11 #link https://review.openstack.org/#/c/333371/ 14:25:22 alexchadin : any update on rally support . 14:25:23 ? 14:25:47 acabot: 1 commit left to get full rally support for watcher 14:25:57 4 are already merged 14:26:36 #info 1 commit left to get full rally support for watcher 14:26:50 #topic Blueprint/Bug Review and Discussion 14:26:55 after that I will update this PS 14:27:03 our current milestone in Newton-3 due August 29th 14:27:10 #link https://launchpad.net/watcher/+milestone/newton-3 14:27:33 our 3 essentials BPs are almost done 14:28:29 gzhai-mobl : did you start working on a spec for "Use a state machine to model Watcher objects lifecycle" ? #link https://blueprints.launchpad.net/watcher/+spec/statemachine-transition-permission 14:29:32 Handle race condition of sync from multiple decision engines should be abandoned (mid-cycle discussion) 14:29:35 not yet. i am on some ceilometer task this week 14:29:37 i will start it soon 14:29:46 gzhai-mobl : ok 14:30:10 could you also abandon this change ? https://review.openstack.org/#/c/339285/ 14:30:20 gzhai-mobl : could you also abandon this change ? https://review.openstack.org/#/c/339285/ 14:30:23 acabot: got it 14:30:39 acabot did you see the discussion about this on the laucnhapad bug comments section? 14:30:58 ok 14:31:06 https://bugs.launchpad.net/watcher/+bug/1598946 14:31:06 Launchpad bug 1598946 in watcher "Initial execution of parallel engines can be racey " [Undecided,In progress] - Assigned to Zhai, Edwin (edwin-zhai) 14:31:28 vincentfrancoise : yes we looked at it during the midcycle 14:31:45 ok good :) 14:31:50 and stated that its not really a bug and we dont have to support it 14:32:12 #action abandon the change https://review.openstack.org/#/c/339285/ 14:32:25 #action gzhai-mobl abandon the change https://review.openstack.org/#/c/339285/ 14:32:56 gzhai-mobl : any update regarding perform scalability test for watcher ? 14:33:20 gzhai-mobl : I mean is the requirements list OK for the deployment team ? 14:33:43 #link https://etherpad.openstack.org/p/watcher-test-environment-specifications 14:34:07 they confirm the configuration 14:34:39 gzhai-mobl : ok they dont need any more details ? 14:34:50 they ask for the scripts 14:35:08 for Monasca deployment ? 14:35:52 dtardivel will be back next week and share scripts 14:36:27 jinquan1 : resize-overload-instance needs an updated description and specs 14:36:39 yes. they didnt deploy monasca before 14:37:03 but they said it should be ok with some help from us 14:37:08 #action dtardivel send Monasca deployment scripts with gzhai-mobl 14:37:08 acabot:yes, will do 14:37:15 In recent weeks, mainly in the company's internal projects, so less time to invest watcher. 14:37:27 i will back to watcher after tommorow 14:37:30 jinquan-tmp : midcycle outputs are clear enough ? 14:38:01 umm, i maybe need discuss with jed 14:38:03 jinquan-tmp : ok no pb, do not hesitate to ask if you need more explanations 14:38:28 jinquan-tmp: yes we can discuss today 14:38:31 if you want 14:38:48 we didn't require cinder, right? 14:39:04 jed56, to late for me today :) 14:39:09 too 14:39:15 jinquan-tmp: ah yes sorry 14:39:33 gzhai-mobl : right, we dont need it 14:39:34 jed56: i will contact you next day 14:39:40 ok? 14:39:47 jinquan-tmp: yes :) 14:40:01 #topic Open discussions 14:40:02 acabot: ok 14:40:04 jed56: thank you very much ! 14:40:17 jinquan-tmp: welcome 14:40:19 Thanks to POC [9] & bench [10] done by vincentfrancoise, we need to decide which metrics back-end we need to support in Watcher 14:40:26 #link https://review.openstack.org/#/c/348892/ 14:40:32 #link https://pad.b-com.com/p/monasca-ceilometer-ceilosca 14:41:00 I really want to avoid having one strategy implemented for each metrics back-end 14:41:42 with the bench, we see that Monasca is really better than telemetry 14:41:59 but Monasca is a nightmare to deploy in a devstack env 14:42:34 so if we want to keep it consistent (telemetry for devstack and monasca for real deployment) 14:42:36 suggest ceilosca considering the performance improvements 14:42:46 we should rely on ceilosca 14:42:55 hvprash : exactly 14:43:08 we have 2 problems with Ceilosca 14:43:27 performances are poor compare to standalone Monasca 14:43:38 and we cannot use groupby is requests 14:43:43 s/is/in 14:44:13 so I would suggest to share our bench results with Monasca team and see if they can improve ceilosca 14:44:23 +1 14:44:34 instead of choosing between Monasca or Telemetry support 14:44:59 internally we were planning to go the ceilosca route 14:45:22 hvprash: whilst trying out ceilosca 14:45:52 jwcroppe edleafe : any thoughts on this ? 14:46:00 I had some tiny behaviour differences between plain old ceilometer and ceilosca 14:46:29 I haven't looked into the differences. Are they significant? 14:46:40 one being when you don't specify a start time 14:46:45 which one for scalability test? monasca 14:46:46 ? 14:46:58 gzhai-mobl : yes monasca 14:47:13 edleafe : differences between telemetry and monasca ? 14:47:21 ceilosca will give you data starting from epoch whereas ceilometer seems to limit it to 1 mont-worth of data 14:47:43 acabot: hmm ... 14:47:52 acabot: yes, as well as ceilometer/ceilosca differences 14:48:00 vincentfrancoise, what behavior differences ? we could discuss after the meeting too. interested in knowing more. 14:48:21 vincentfrancoise: so if you specify a start, they are the same? 14:48:40 edleafe: yes 14:48:54 that's why I said tiny differences 14:48:59 ah ok 14:49:16 watcher strategies we not giving any start time until now 14:49:33 and it was working although slowly using Ceilometer 14:49:40 I would strive to keep the interface consistent, and work to make ceilosca as close to ceilometer as necessary 14:49:55 This keeps the Watcher side simple, and improves ceilosca 14:50:04 +1 14:50:10 but when I switched to Ceilosca, it pretty much broke Watcher because it was returning to much data with zeros everywhere 14:50:13 +1 14:50:27 edleafe: this is the idea of ceilosca but ceilosca have some drawbacks 14:51:19 #action acabot vincentfrancoise share bench results with Monasca team and ask for possible improvements of ceilosca 14:51:21 vincentfrancoise: so it sounds like ceilosca still needs a lot of work 14:51:32 edleafe: yes that's my point 14:51:36 vincentfrancoise: I wouldn't mangle Watcher logic to compensate 14:51:38 edleafe: +2 14:52:30 I think using ceilosca give prefs as good as using ceilometer ;-) 14:52:39 s/prefs/perfs 14:53:02 acabot: it was still 8 times slower 14:53:07 So unless you want to add a driver (translation) layer to the interface to telemetry, I'd standardize on ceilometer and help the ceilosca folks improve their product 14:53:40 edleafe : I want to avoir the driver layer ! 14:53:46 s/avoir/avoid 14:53:56 acabot: yeah, so would I! 14:54:21 ok lets try to sync with ceilosca team and see what happened 14:54:48 any other topic to discuss in the last 5 minutes ? 14:55:30 edleafe: I agree 14:55:33 hvprash : I will send the email for promoting you as core on watcher-specs next week 14:55:51 thx acabot 14:56:27 hvprash : do you know if michaelgugino will be available next week to work on ansible scripts with dtradivel ? 14:56:54 i will check and confirm. i think he should be available 14:57:04 great thx 14:57:19 sure 14:57:45 ok so thank you and have a good day/night 14:58:16 bye 14:58:18 bye 14:58:19 bye 14:58:21 Bye 14:58:21 bye 14:58:24 #endmeeting