09:01:52 #startmeeting watcher 09:01:53 Meeting started Wed Aug 3 09:01:52 2016 UTC and is due to finish in 60 minutes. The chair is jed56. Information about MeetBot at http://wiki.debian.org/MeetBot. 09:01:54 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 09:01:56 The meeting name has been set to 'watcher' 09:02:09 Hello 09:02:13 hi 09:02:32 o/ 09:02:58 o/ 09:03:31 Agenda for today #link https://wiki.openstack.org/wiki/Watcher_Meeting_Agenda#08.2F03.2F2016 09:03:42 hi 09:04:12 #topic Announcements 09:04:45 #info I will send a mail to promote alexchadin as a watcher core contributor today 09:05:12 hi 09:05:19 any other announcement ? 09:05:50 some parts of watcher support in rally has been merged 09:06:19 2 patches is waiting for new PS 09:06:23 can you share with us the links ? 09:06:29 yeap, wait a sec 09:07:47 https://review.openstack.org/#/q/openstack/rally+owner:%22Alexander+Chadin+%253Ca.chadin%2540servionica.ru%253E%22 09:08:09 #info some parts of watcher support in rally has been merged 09:08:18 #link https://review.openstack.org/#/q/openstack/rally+owner:%22Alexander+Chadin+%253Ca.chadin%2540servionica.ru%253E%22 09:08:29 thanks 09:08:33 #topic Review Action Items 09:08:38 #topic watcher specs 09:09:05 The Graph model describing virtual and physical elements in a data center blueprint has to be reviewed by the cores because the development will start soon. 09:09:21 #info https://review.openstack.org/#/c/298891/ need to be merged 09:09:37 #topic watcher 09:10:01 The blueprint 'cluster-model-objects-wrapper' has been merged so we can start to develop 'Graph model' blueprint. 09:10:08 The first part of the scoring-module has been merged and also the blueprint 'bp/persistent-audit-parameters'. 09:10:15 Congratulations to all those who worked on these blueprints ! 09:11:15 +1 :) 09:11:23 #info 'cluster-model-objects-wrapper' first part 'scoring-module' persistent-audit-parameters are available in the watcher trunk 09:11:42 As we have merged many things, I would like that everybody take some time to test Watcher on his environment. 09:11:57 +1 09:12:23 will do 09:12:35 Who could take some time to test watcher ? 09:12:43 I'll do 09:12:58 I am trying out everything now 09:13:00 +1 09:13:22 #action diga alexchadin hanrong gzhai test watcher on his own infrastructure 09:13:31 +1 09:13:40 I will also do some testing as part of testing scoring module 09:13:56 This is better that each developer of the available watcher strategies take some time 09:14:16 this is the more "complex" part to evaluate right now 09:14:31 +1 09:14:49 for example if alexchadin you can test https://github.com/openstack/watcher/blob/master/watcher/decision_engine/strategy/strategies/workload_stabilization.py 09:14:53 unit tests on strategy do not cover all possible scenarios 09:15:02 vincentfrancoise: +2 :) 09:15:55 jed56: ok, once I discuss with you audit scope a little bit later:) 09:16:06 alexchadin: yes 09:16:19 vincentfrancoise: has started to evaluate monasca 09:16:28 https://review.openstack.org/#/c/348891/ 09:16:37 #info vincentfrancoise: has started to evaluate monasca 09:16:51 vincentfrancoise: Can you share with us some results ? 09:17:11 jed56: I will be full-time available since 8 august, is it okay for you? 09:17:40 so here is a pad that show some of the results I had while playing around with monasca: https://pad.b-com.com/p/monasca-ceilometer-ceilosca 09:17:44 no full-time for me enough time to help you ;) 09:18:11 vincentfrancoise: how about current status of monasca? 09:18:21 Is it ok for scalability test? 09:18:26 #link monasca evaluation results https://pad.b-com.com/p/monasca-ceilometer-ceilosca 09:18:35 so basically I pushed a monasca helper that will be available for strategy developers 09:18:51 using influxDB as storage with monasca 09:19:13 we can see that the metrics queries are much faster 09:19:22 so need change strategy's code? 09:19:26 It's great! 09:19:43 so there are 3 approaches 09:20:26 with the ceilometer one, it took 18s to execute the basic_consolidation strategy 09:20:59 but with a tweak in the query formatting I could actually reduce it down to 4 seconds 09:21:28 see https://review.openstack.org/#/c/348891/2/watcher/common/ceilometer_helper.py 09:21:58 which means that we can still use ceilometer with "acceptable" performances 09:22:19 i see. limit the query. 09:22:29 but compared to monasca, we can get down to .5 seconds with the same configuration 09:22:50 Fantastic! 09:23:10 total strategy execution time? 09:23:29 yes that includes the time spent building the cluster data model (nova queries) 09:23:44 #info performances => ceilometer 4 seconds versus monasca 0.5 seconds 09:24:35 so, we have to decide if we keep the both ( ceilometer or monasca) 09:24:40 because the queries are not the same 09:24:56 gzhai: if you look at line 11 and 12 in the pad I explain the difference 09:25:13 jed56: +1 09:25:28 #info we have to decide if we maintain ceilometer and monasca in watcher 09:25:54 vincentfrancoise: the strategy code can handle both, right? 09:26:11 https://review.openstack.org/#/c/348892/ shows how much I had to change basic_consolidation to make it work using monasca 09:27:29 I forgot to say that there is new spec https://review.openstack.org/#/c/348049/ from joe 09:27:32 what is dimensions? 09:27:47 gzhai: yes but there is not abstraction with the 2 so the strategy will have to be explicit on whether it uses ceilometer or monasca 09:28:01 vincentfrancoise: +2 09:28:01 alexchadin: dimensions are pretty much filters 09:28:28 we need rewrite all existed strategies for monasca? 09:28:39 Keep both ceilometer and Monasca copy? 09:28:45 if we decide to move away from ceilometer: yes 09:29:16 gzhai: that's a question that will have to be answered at some point 09:29:54 that's why I pushed the basic_consolidation updates as a [PoC] 09:30:03 got it. thanks. 09:30:23 jed56: can we implement the celometer & monasca both 09:30:43 I think this way, we are giving more options to the user 09:31:16 as vincentfrancoise says this is possible but this is not easily 09:31:34 s/easliy/easy 09:31:36 todo 09:31:44 yes, okay 09:31:50 the best thing that can be done would be to add a strategy parameter asking which metrics backend to use 09:31:52 diga: I agree the best solution is to have the boths 09:32:03 vincentfrancoise: I agreee 09:32:09 jed56: yep 09:32:24 #topic Blueprint/Bug 09:32:31 https://review.openstack.org/#/c/343135/ has been merged 09:32:45 oops, need a new patchset :) 09:33:27 but https://review.openstack.org/#/c/342594/ has been merged 09:34:08 any others bugs ? 09:34:17 #topic open discussion 09:34:56 I have not open discussion. 09:34:59 :) 09:35:01 we have shifted meeting from 1600 UTC to now ?? 09:35:14 is this the fix timing of meeting ? 09:35:27 this is changing every week 09:35:42 alternative weeks 09:35:43 ok 09:35:47 we meet weekly on Wednesdays at 14:00 UTC on even weeks, 9:00 UTC on odd weeks 09:36:00 okay 09:36:03 got it 09:36:21 htop 09:36:27 100% 09:36:29 100% 09:36:34 :D 09:36:43 sorry guys, wrong window :) 09:36:52 bye 09:36:55 hahah 09:36:58 bye 09:37:01 logout 09:37:05 bye 09:37:09 #endmeeting