14:01:02 <andreykurilin_> #startmeeting Rally 14:01:03 <rallydev-bot2> [From Gitter] astarove : Hi 14:01:03 <openstack> Meeting started Mon Mar 13 14:01:02 2017 UTC and is due to finish in 60 minutes. The chair is andreykurilin_. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:01:04 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:01:06 <openstack> The meeting name has been set to 'rally' 14:01:39 <hai_shi> hi 14:02:40 <andreykurilin_> we do not have meetings for a long time. It is time to return back that event 14:02:54 <andreykurilin_> I have several items to discuss 14:03:04 <andreykurilin_> #topic Status of Rally 0.9.0 14:03:14 <andreykurilin_> Rally 0.9.0 is our new release 14:04:02 <andreykurilin_> We merged all features required for that release 14:04:13 <andreykurilin_> but they produced some "bugs" :) 14:05:18 <andreykurilin_> our new cleanup mechanism doesn't cover the case of auto-created resources :( 14:05:59 <andreykurilin_> for example in case of booting VM with auto_assign_nic=True, Nova will assign floating IP which we don't cleanup 14:06:17 <andreykurilin_> such resources doesn't match our name patterns and it is hard to identify them 14:07:10 <andreykurilin_> I'll try to find the way to discover and remove such resources 14:07:19 <andreykurilin_> it is only one blocker for Rally 0.9.0 14:07:57 <hai_shi> Do you find some better way to resolve it? 14:08:21 <andreykurilin_> hai_shi: not yet :) 14:08:44 <andreykurilin_> we should not revert your change, since it is a big step forward 14:09:11 <andreykurilin_> but should add support for another method of discovering resources 14:09:28 <andreykurilin_> hope, today I'll produce a PoC 14:10:34 <hai_shi> can we update those resources' name? 14:11:48 <hai_shi> or before use it, we should tag them. 14:12:07 <andreykurilin_> hai_shi: no. for example, we do not have access for creating params of floating-ip created while "booting VM with auto_assign_nic=True", since that resource is not created by Rally itself. Nova-API does it automatically 14:15:00 <andreykurilin_> ok, let's move to the next topic 14:15:09 <andreykurilin_> #topic plans for the next release 14:15:54 <andreykurilin_> Version 1.0.0 was reserved long time ago for the release of Rally that will support not-only OpenStack platforms 14:16:05 <andreykurilin_> and it looks like the time is come 14:16:33 <andreykurilin_> I hope that tohin will finish refactor of credentials and deployment during the next release 14:16:36 <andreykurilin_> :) 14:17:20 <chenhb_> Hi 14:17:22 <andreykurilin_> another things that should be finally merged: glance and cinder "services".. 14:17:30 <andreykurilin_> hi chenhb_! 14:17:36 <hai_shi> hi 14:18:05 <andreykurilin_> also, RaaS stuff should be started and atomic actions refactoring 14:18:13 <andreykurilin_> actually, we have a lot of items for the next release 14:18:15 <andreykurilin_> :) 14:18:41 <hai_shi> a big step! 14:19:38 <chenhb-> ping 14:19:47 <andreykurilin_> chenhb- pong 14:20:23 <andreykurilin_> So let's schedule Rally 1.0.0 release to the middle-end of April. 14:21:23 <hai_shi> No problem. 14:21:32 <andreykurilin_> cool 14:22:13 <chenhb-> copy 14:22:17 <andreykurilin_> ok, so list with items to share is finished 14:22:23 <andreykurilin_> *my list 14:22:27 <andreykurilin_> any topics to discuss? 14:22:54 <andreykurilin_> #topic Free discussion 14:23:09 <hai_shi> hi, andrey. Do we have some policies about disater-cleanup? 14:23:51 <hai_shi> I see we have many items to do in bp of cleanup. 14:24:06 <chenhb-> about service module, do we need to generate name in service? 14:24:23 <chenhb-> or move to scenarios 14:25:56 <andreykurilin_> hai_shi: policies? so it is one more "nice to have" feature :) I need to revise spec for it (I do not remember work-items from it), but it looks like we need just to implement entry point for it and call cleanup mechanism with proper task uuid 14:25:58 <hai_shi> i see generate_name in scenarios' util serveral times. 14:27:07 <hai_shi> Ok, I just see some items need to do in bp. 14:27:38 <andreykurilin_> chenhb-: In most cases scenario(or context) doesn't interested in name of resource, so for simplification and unification, services should generate names by default(if name is not transmitted by scenario) 14:28:19 <hai_shi> Why we should transmitted name by scenarios? 14:29:33 <hai_shi> user transmited name and we didn't use it~~ 14:30:01 <chenhb-> If user use custom name,our new cleanup can clean them? 14:30:25 <andreykurilin_> when I said scenario, I mean python code. Setting names from task (via JSON/Yaml) should be restricted 14:31:17 <andreykurilin_> chenhb-: no, our cleanup mechanism will not remove resources with custom names 14:31:59 <hai_shi> https://review.openstack.org/#/c/432138/3/rally/plugins/openstack/scenarios/cinder/utils.py@430 14:33:39 <andreykurilin_> @hai_shi: we should restrict such scenarios. For checking feature of "updating name", we should accept boolean argument "update_name" via task, and call generate_random_name by in python scenario 14:33:42 <hai_shi> andrey: What is your idea about this test? it is transmited name in scenario. 14:33:47 <fungi> mordred: http://lists.openstack.org/pipermail/openstack-infra/2017-March/005240.html seems to suggest that manage-projects isn't updating. any chance you want to take a look? 14:34:00 <hai_shi> ok, copy that. 14:34:08 <chenhb_> ok, keep name in python code 14:34:14 <fungi> oops, sorry, wrong channel :/ 14:34:27 <andreykurilin_> fungi: np :) 14:36:32 <andreykurilin_> chenhb: I see a lot of "reconnection" messages for you. If you have problems with IRC, you can use our gitter channel for meetings, it has configured sync-bot too 14:37:26 <chenhb_> yep, nothing from me now: ) 14:37:45 <andreykurilin_> about cleanups and names: Since we merged "new cleanup", let's check all scenarios and restrict setting names via tasks 14:38:18 <chenhb_> copy 14:38:26 <hai_shi> ok 14:38:37 <andreykurilin_> anything else? 14:39:19 <hai_shi> small question 14:39:54 <andreykurilin_> sure 14:40:28 <hai_shi> when use auto_assign_nic, we would use the same network again and again due to the random function. 14:40:49 <hai_shi> But it is not a good way to test. 14:44:06 <andreykurilin_> hai_sho: it is not random at all:) picking random networks tries to balance the load between available networks. if you setup networks context to create N networks, where N is a number of iterations, each iteration will use separate network 14:44:11 <andreykurilin_> *hai_shi 14:44:33 <hai_shi> in this lines https://github.com/openstack/rally/blob/master/rally/plugins/openstack/scenarios/neutron/utils.py#L296-L297 14:45:02 <andreykurilin_> ok, so it is not about auto_assign_nic at all 14:45:24 <andreykurilin_> https://github.com/openstack/rally/blob/master/rally/plugins/openstack/scenarios/nova/utils.py#L98-L106 14:45:29 <hai_shi> ok, i forget details:( 14:45:33 <andreykurilin_> that function is used with booting vms 14:47:01 <hai_shi> it looks better. 14:48:18 <andreykurilin_> as for neutron scenarios, some of them use https://github.com/openstack/rally/blob/master/rally/plugins/openstack/scenarios/neutron/utils.py#L296-L297 and yes, it makes unfair situation. There are two possible solutions- adopt _get_or_create_network to do not use random (as it done in _pick_random_nic) or abandon ability to use pre-created networks 14:50:20 <hai_shi> thanks, andrey. I have no other problem. 14:50:43 <andreykurilin_> any other topics to discuss? 14:51:03 <andreykurilin_> ok, let's finish 14:51:18 <andreykurilin_> thanks to everyone! 14:51:19 <andreykurilin_> bye 14:51:22 <andreykurilin_> #endmeeting