14:02:54 #startmeeting rally 14:02:55 Meeting started Mon Jul 27 14:02:54 2015 UTC and is due to finish in 60 minutes. The chair is boris-42. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:02:56 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:02:58 The meeting name has been set to 'rally' 14:03:15 rvasilets: amaretskiy oanufriev ping 14:03:20 hi 14:03:24 \o 14:04:18 hi 14:04:54 Hi 14:05:12 :) 14:07:33 boris-42: here 14:08:00 o/ 14:08:04 #topic Make Rally framework independent from OpenStack 14:08:33 As far as you know I am working quite a lot on this task 14:08:44 https://review.openstack.org/#/c/203911/ 14:08:56 ^ this patch resolves one more place that blocks this effort 14:09:36 e.g. Base runner should know to content of Context object to prepare scenario context 14:09:41 which is terrible 14:10:02 after this patch each context plugin is able to transform context object for scenario 14:10:24 which allows us to make runner independent of context obj 14:11:33 e0ne: rvasilets amaretskiy redixin oanufriev kun_huang yingjun ^ any questions 14:11:46 no 14:11:54 everything is clear? 14:12:06 no 14:12:06 nope, I will watch it 14:12:28 boris-42, no. Interesting patch which need more patient from me 14:13:20 watch it later 14:13:25 okk 14:13:26 i'll review it later today 14:13:46 next topic 14:13:49 #topic [stpierre] create_and_list_ (e.g., create_and_list_image) or create_and_list_ (e.g., create_and_list_users)? We use a mix of both, and should pick one and standardize. 14:14:48 +1 for plural 14:15:21 +1 for plural 14:15:27 imho it should be obvious 14:15:42 yes, +1 for plural 14:15:45 who want to make test for this? 14:16:00 redixin: pboldin LOL=) 14:16:07 rofl 14:16:15 he like regexp 14:16:47 lets vote) +1 from me for pboldin) 14:16:56 LOL 14:18:45 e0ne: ^ 14:18:48 skraynev: ^ 14:19:29 boris-42: +1 from me about using plural everywhere 14:19:39 +1 14:19:41 kun_huang, ^ 14:19:48 temujin, ^ 14:20:02 +1 14:20:13 +1 for plural 14:20:14 plural 14:20:14 +1 for plural 14:20:15 #agreed we should use plural everywhere and make hacking rule for this 14:20:31 #topic 14:20:33 #topic [e0ne] verify performance with Rally: compare performance of scenarios after 2 runs 14:20:49 e0ne: this is question to amaretskiy * 14:20:56 in cinder, we want to get in in infra 14:21:09 e0ne: oh you mean to do this in gates? 14:21:13 yep 14:21:42 e0ne: ok then it's questions to me 14:21:50 :) 14:21:52 boris-42: I'm ready to start implementing that after having patches related to new reports classes merged - these patches are ready for review (having all +1 from gates) 14:22:00 e0ne: compare the performance with new patch and without new patch in gate? 14:22:02 amaretskiy: it's a bit different topic 14:22:12 boris-42: okay 14:22:23 boris-42: I can start this separately 14:22:34 kun_huang: yes. but i understand that we could get false-negative results 14:22:39 amaretskiy: that will be definitely useful 14:22:46 e0ne: yep 14:22:56 boris-42: okay, starting work on this patch... 14:23:05 e0ne: because gate jobs are run on different VMs 14:23:29 boris-42: you're right :( 14:23:39 e0ne: and some of VM are 2 times more powerful than other 14:23:49 e0ne: so I hade some ideas about this 14:24:07 e0ne: boris-42 is it possible to do this in 3rd CI? 14:24:16 e0ne: e.g. running few benchmarks to measure CPU/IO/MEM/Ne 14:24:28 kun_huang: that is possible 14:24:49 e0ne: and then using this number normalize results 14:25:01 e0ne: but it's quite hard task imho 14:25:16 boris-42: sure 14:25:29 e0ne: boris-42 https://01.org/sites/default/files/styles/project_screenshot/public/lkp-homepage.gif?itok=PHTL0syr 14:25:40 boris-42: and it will be implemented maybe only in N release 14:25:47 in linux kernel, they do the job like this 14:26:02 e0ne: yep like that 14:26:20 kun_huang: so the thing is that in OpenStack they want to do this on every patch 14:26:22 kun_huang: in CI 14:26:52 boris-42, we could find after some running of task the weakest vm. And start to use it like measure value) That try to find const that athoer vm have relitevaly to 'basic' vm. And try to divide avery value in proper graphic by this const maybe 14:27:18 rvasilets: that won't work 14:27:33 rvasilets: we need to find the way to normalize results 14:27:34 it would be work fine of we have some permanent set of vms 14:27:52 rvasilets: nope it won't work because you can get merged patch 14:28:02 rvasilets: that destroys performance (e.g. 2 times) 14:28:20 rvasilets: after that you won't be able to merge anything because all other VMs won't be so powerful 14:28:25 and you will block gates 14:28:41 as well the idea is to have this job fully automatic 14:28:59 boris-42: +1 14:29:05 e.g. if performance was improved it should automatically find new baseline 14:30:48 boris-42, agree 14:30:56 okay so someday we will work on this 14:31:07 but there are simpler and more important tasks at this point 14:31:27 as well we should prepare performance trends reports 14:31:32 which amaretskiy is already working on 14:31:34 e0ne: ^ 14:31:50 sounds good 14:31:53 great! 14:33:52 #agreed We will finish this task someday 14:34:16 #topic [rvasilets] Replace depraceted oslo.db code 14:34:24 rvasilets: so^ 14:35:31 #link https://review.openstack.org/#/c/205530/ Why looking at logs I found that https://github.com/openstack/oslo.db/blob/master/oslo_db/sqlalchemy/enginefacade.py#L830 14:35:50 We have depraceted usage of oslo.db at code 14:36:00 so I decide ro write a small patch 14:36:21 if you like this idea please review it 14:36:22 eom 14:36:46 rvasilets: yep we should do this 14:38:23 it seems reasonable for me 14:38:43 that's the suggestion from oslo 14:39:44 +1 14:39:56 btw boris-42 rvasilets e0ne does gate-rally-dsvm-verify block everyone now? 14:40:13 kun_huang: not sure 14:40:23 #agreed gate-rally-dsvm-verify looks broken 14:40:26 kun_huang: I had to do the stuff and didn't track rally for last 4 days 14:40:34 #topic gate-rally-dsvm-verify looks broken 14:40:35 нуы 14:40:36 yes 14:40:41 they a blocked 14:40:44 okay I will take a look today at it 14:40:48 and try to unblock it 14:40:54 from 23 tp 24th 14:41:03 nigth 14:41:04 https://review.openstack.org/#/c/198877/4 14:41:09 looks like by this patch 14:41:17 I have talk with redixin 14:41:54 and he thin that problem that we don;t have permissions at absolute paths 14:42:33 rvasilets, did you try to install tempest into venv? 14:42:56 I tried but found some other issues 14:43:18 and I don't now what is they mean) 14:43:22 rvasilets: uh oh 14:44:26 Do it install only with cloud or without too? 14:45:24 Devstack have broken for me now. And without cloud I got following http://paste.openstack.org/show/405924/ 14:47:48 rvasilets: ok guys 14:47:58 we will disucss this in rally chat 14:48:07 #topic Open discussion 14:48:07 ok 14:48:14 do we have something to discuss? 14:48:36 none from me 14:48:42 nothing special from me 14:48:45 No. Nothing significant) 14:50:49 okay let's finish meeting 14:50:55 #endmeeting