17:01:44 #startmeeting rally 17:01:44 Meeting started Tue Apr 7 17:01:44 2015 UTC and is due to finish in 60 minutes. The chair is boris-42. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:01:45 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 17:01:47 The meeting name has been set to 'rally' 17:01:54 oanufriev: msdubov meteorfox ping 17:02:22 o/ 17:02:55 meteorfox: hi htere 17:05:14 amaretskiy: ping 17:05:16 boris-42 hi! 17:05:19 hi 17:05:32 hi 17:06:44 boris-42: Here :D 17:07:02 rook: nice) 17:07:09 e0ne: around? 17:07:22 boris-42: yep 17:07:34 okay I believe we can start?) 17:07:44 rally meeting? 17:07:49 e0ne: yep 17:08:04 #topic Rally stackforge->openstack 17:08:23 okay seems like most of TC are already ready to add Rally to OpenStack 17:08:33 https://review.openstack.org/#/c/169357/ 17:08:50 ^ I had to rebase and rephrase it a bit but in history most of TC are ok 17:08:59 awesome! congrats 17:09:00 awesome 17:09:01 so that's actually very nice news=) 17:09:47 outstanding! 17:10:12 boris-42:congrats! 17:10:58 msdubov: rook meteorfox e0ne so today in 3 hours they will make final decision 17:11:06 there will be TC meeting in this chat 17:11:40 cool, I'll be around. I hope it gets in 17:12:33 ok 17:13:07 #topic Online Bash competition & making Rally start fast 17:13:14 So we have deep issues with Rally 17:13:23 it takes about 600-700ms to start rally 17:13:35 boris-42: do you meant completion? 17:13:46 rally ta 17:14:34 meteorfox: ^ 17:14:36 ok, yeah. then 'completion' it is. I was confused, I thought there was some kind of contest or something 17:14:46 meteorfox: yep=) 17:14:58 completion* 17:15:06 so we can't make it online* 17:15:18 it has to be hardcoded and not generate online 17:15:23 to fix this I created tool 17:15:34 https://github.com/boris-42/profimp 17:15:46 ^ so we can analyze what is imported so long and optimize rally imports 17:16:34 cool! 17:17:16 meteorfox: so it allows to trace imports and some of libs already used it 17:17:29 meteorfox: rootwrap and keystone to make their imports faster 17:17:52 boris-42: this will be useful for other projects, I wasn't aware this existed. thanks man! 17:18:41 meteorfox: hehe=) 17:18:43 meteorfox: star it=) 17:18:50 just did 17:18:56 okay so we are going to imporve rally 17:19:02 make it start fast 17:19:07 and implement online bash completition 17:19:37 is there an example of how to use it? 17:19:44 meteorfox: readme 17:19:50 thought I saw a example on the main page? 17:20:05 meteorfox: yep just scroll page 17:20:06 right, I meant more as a library 17:20:23 meteorfox: ?) 17:20:51 boris-42: nevermind, I can figure out 17:21:34 meteorfox: you mean use profimp as a lib? 17:21:44 meteorfox: from profimp import main 17:21:51 boris-42: yeah, I saw I can use trace_module() 17:22:02 main.trace_module() 17:22:07 thta's all that you need to know=) 17:22:15 yep yep 17:22:20 it will return json 17:22:42 actually no it will return root instance of TracePoint 17:22:49 it has to_dict() method 17:23:04 nifty 17:23:33 okay 17:23:40 let's move to next topic 17:23:47 #topic Important changes 17:24:18 we restore abbility to use boot_runcommand_delete with fixed ip 17:24:22 amaretskiy: thanks ^ 17:24:28 Add new tutorial 17:24:40 https://rally.readthedocs.org/en/latest/tutorial/step_8_task_templates.html 17:25:09 Now SSL configuration is per deployment not per rally.conf 17:25:18 and rally.conf one is deprecated 17:25:45 As well OS_ENDPOINT is recognized in rally deployment --fromenv 17:26:07 As well we merged bunch of new benchmarks 17:26:13 So that is all for this week 17:27:22 Next topic?) 17:27:24 any questions? 17:28:35 meteorfox: rook & 17:28:36 ^ 17:28:43 nope 17:28:48 okay 17:28:51 has anyone played much with rps? 17:29:02 rook: so I played a bit 17:29:12 rook: but seems like we have some issues with it 17:29:37 I personally havent, but members who I work with have seen some problems... where Rally doesn't catch a guest went into ERROR state (vmtask specifically). 17:29:43 boris-42: ah ok. 17:29:51 boris-42: I was going to have him open a BZ 17:30:00 rook: oh nope 17:30:19 rook: I was talking about issues related with proper load generation 17:30:21 rook: not such 17:30:29 ITER: 0 END: Error TimeoutException: Rally tired waiting for unicode u'10.35.180.25': to become current status NONE 17:30:34 ^ example boris-42 17:30:38 rook: so this is ok 17:30:53 Guest was in ERROR state, but Running 17:30:55 rook: it can't ping VM it is not related to runner 17:31:17 rook: what do you mean by Running?) 17:31:56 boris-42: Status was ERROR, Power State was Running 17:32:09 | c31dff3b-a03b-41e2-9799-cb9cc35a79ea | sriov | ACTIVE | - | Running | net=10.0.0.7 | 17:32:15 ^ Example where things are "good" 17:32:30 rook: ITER: 0 END: Error TimeoutException: Rally tired waiting for unicode u'10.35.180.25': to become current status NONE 17:32:30 17:32:43 ^ This means that Rally tired wating to ping VM 17:32:59 Right, but that guest was in ERROR state... 17:33:11 I personally haven't investigated this much 17:33:19 i was curious if others have seen where rps causes failures. 17:33:30 rook: it can't be related to rps 17:33:43 rook: the bug may be in vm task boot and run command 17:33:46 rook: scenario 17:33:59 possibly... I need to play with it some 17:34:06 rook: ok 17:34:23 let's move to next topic 17:34:33 #topic Release 0.0.3 is coming 17:34:46 I am going to cut on this week new release 17:35:02 I will try to finish support of existing users before cut of new releae 17:35:56 rook: oanufriev amaretskiy msdubov ^ 17:35:59 meteorfox: ^ 17:36:03 anything?) 17:36:46 nope, the comments I had regarding the 'sigma test' for the outliers, I left them in the patch itself. I think it should be discussed there 17:37:02 msdubov: did you see those comments? 17:37:04 meteorfox: ^ 17:37:10 boris-42 What else should be completed before 0.0.3? 17:37:23 msdubov: I believe only existing users 17:37:31 but they can be even done after 17:37:45 as well this patch is interesting 17:37:46 https://review.openstack.org/#/c/170160/ 17:38:23 any link we have for 0.0.3 items? 17:38:44 boris-42:Yes, I saw it. Maybe also "task abort" in 0.0.3? 17:39:19 pradeep: not yet but let we discuss this in next topic 17:39:29 msdubov: heh not sure if we finish it before existing users and get merged 17:39:31 sure 17:39:39 okay let's move to next topic 17:39:50 #topic How to make simple to understand what is important for next release 17:39:57 boris-42: may be we'll complete 2 more scenarios for cinder this week 17:40:03 e0ne: ok 17:40:08 one is ready for review 17:40:10 So I have idea 17:40:24 it was actually redixin 17:40:28 idea 17:40:33 To make one blueprint 17:40:37 in launchpad 17:40:53 "important-for-next-release" and add it to commit messages 17:41:09 so we will have page with all patches that are important-for-next-release 17:41:19 msdubov: e0ne meteorfox rook pradeep ^ 17:41:27 interesting... 17:41:29 e0ne: regarding to cinder benchmarks 17:41:40 e0ne: okay we will take a look 17:41:53 #link https://blueprints.launchpad.net/rally/+spec/important-for-next-release 17:41:55 boris-42: thanks! 17:42:03 ^ So actually I created such BP already 17:42:06 but we are not using it 17:42:08 ye 17:42:10 yet* 17:42:34 boris-42: I like this idea, redixin thanks for this 17:43:10 boris-42: its nice 17:43:43 idea 17:43:45 okay then after cutting next release we will start adding this bp to all important patches 17:46:47 any other comments? 17:47:12 * rook needs to catch up! ;) 17:47:21 boris-42: Will we use only one bluepirnt? 17:47:27 msdubov: yep 17:47:28 boris-42:Maybe better one BP for each release? 17:47:35 msdubov: nope 17:47:53 msdubov: because if decided to not merge important for release 0.0.x in 0.0.x 17:47:59 we will need to update commit message 17:48:08 boris-42: I see 17:48:28 msdubov: as well I am not sure that it will be easy to predict version of rally always 17:51:27 okay let's move 17:51:32 #topic open dicussion 17:51:49 amaretskiy: msdubov meteorfox rook pradeep anything to discuss?) 17:52:17 Since I am behind on things, is there any work on creating a Rally API ? 17:53:29 rook: so yep 17:53:44 rook: here is rally api (as a lib) https://github.com/stackforge/rally/blob/master/rally/api.py 17:53:58 rook: and we are going to add more methods like list and show to it 17:54:13 rook: and remove rally.objects usages from rally.cli 17:54:25 * rook will check it out. 17:54:27 thanks boris-42 17:54:29 rally.cmd* 17:54:53 rook: so basically on top of rally.api we will create a tinny translator HTTP-2-API 17:55:03 rook: so we will have as well rally as a service 17:58:07 rook: okay 17:58:24 so let's finish meeting 17:58:28 see you 17:58:32 #endmeeting