14:07:18 <boris-42> #startmeeting Rally
14:07:19 <openstack> Meeting started Mon Mar 28 14:07:18 2016 UTC and is due to finish in 60 minutes.  The chair is boris-42. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:07:20 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:07:22 <openstack> The meeting name has been set to 'rally'
14:07:24 <andreykurilin> nice
14:07:27 <boris-42> andreykurilin: hm?
14:07:30 <andreykurilin> o/
14:07:33 <anteaya> if the bot left or was disconnected at some point often the topic is not current
14:07:39 <rvasilets> o?
14:07:45 <boris-42> okay so it works
14:07:48 <boris-42> anteaya: thanks
14:07:54 <anteaya> after the meeting if someone could ping someone in the -infra channel they can refresh the topic
14:07:58 <anteaya> boris-42: welcome
14:08:10 <andreykurilin> anteaya: got it, thanks:)
14:08:19 <boris-42> so
14:08:28 <andreykurilin> today we have an agenda
14:08:38 <andreykurilin> \o/
14:08:40 <amaretskiy> finally :)
14:08:41 <boris-42> #topic I finally did something for rally
14:08:47 <ikhudoshyn> ))
14:08:57 <boris-42> so basically there is a bug in Rally code
14:08:58 <andreykurilin> boris-42: really?
14:09:16 <boris-42> we are not calculating properly load duration
14:09:23 <rvasilets> no way)
14:09:25 <boris-42> I fixed that
14:10:04 <boris-42> so now load duration is calulated as a duration between first iteration start and biggest (started_at + duration) iteration
14:10:15 <redixin> we need to fix the rest https://bugs.launchpad.net/rally/
14:10:26 <boris-42> amaretskiy: ^ this should remove artefacts on reports
14:10:34 <boris-42> redixin: yep
14:10:47 <amaretskiy> boris-42: great fix
14:10:57 <boris-42> btw it's simple
14:11:20 <boris-42> so next topic is about pain that happen after that fix
14:11:36 <boris-42> #topic Rally / MGMT network / Pain
14:11:58 <boris-42> so guys seems like it's very hard to use rally when you don't have access to mgmt network
14:12:46 <boris-42> so there are 2 ways to address this
14:13:02 <boris-42> one is to finish faster deployment & verification refactoring
14:13:20 <boris-42> another is to finish work on rally as a service and run rally as a s regular service
14:13:42 <redixin> i know one more way
14:13:49 <amaretskiy> run rally on separate nod and use raas - the best way imho
14:13:52 <redixin> to create python-horizonclient
14:13:57 <amaretskiy> *node
14:14:04 <boris-42> redixin: that will work VERY VERY SLOW
14:14:16 <redixin> that was a joke
14:14:26 <boris-42> redixin: btw people is working on that
14:14:30 <boris-42> =)
14:14:37 <redixin> -_-
14:15:17 <andreykurilin> boris-42: I'll try to finish aas spec asap
14:16:21 <boris-42> andreykurilin: thanks
14:16:50 <boris-42> okay
14:17:16 <boris-42> #topic [amaretskiy] Balancing users: maybe do not support random choice at all, just switch to balancing silently?
14:17:41 <boris-42> amaretskiy: so I think that we neeed to support random choice
14:17:48 <amaretskiy> okay
14:17:54 <boris-42> amaretskiy: however I do agree that we can switch to round robin
14:17:58 <boris-42> amaretskiy: by default
14:18:10 <amaretskiy> this looks reasonable to me
14:18:27 <amaretskiy> balancing is what users expect
14:18:40 <boris-42> amaretskiy: so I do not think that it would be hard to support both
14:18:47 <amaretskiy> okay
14:18:52 <boris-42> amaretskiy: and add parameter in context for that
14:19:00 <amaretskiy> i was in doubt so I decided to add this topic
14:20:33 <andreykurilin> boris-42: lets discuss https://review.openstack.org/#/c/229896/
14:20:50 <boris-42> andreykurilin: we already did, no?
14:21:03 <andreykurilin> it should not be blocker for distributed runner
14:21:12 <andreykurilin> amaretskiy: am I right?
14:21:30 <amaretskiy> not blocking
14:21:38 <boris-42> amaretskiy: andreykurilin so there is way to implemnet it
14:21:43 <boris-42> amaretskiy: andreykurilin that won't block
14:21:44 <amaretskiy> because each d-runner has its own index
14:21:59 <amaretskiy> so we can add shift by d-runner index
14:22:00 <boris-42> amaretskiy: andreykurilin so i think i remember the problem here
14:22:15 <boris-42> amaretskiy: andreykurilin we are not passing iteration index to context
14:22:22 <amaretskiy> no no
14:22:26 <amaretskiy> no titeration index
14:22:32 <amaretskiy> thi spatch uses RAMInt
14:22:42 <amaretskiy> so in case if we have d-runners
14:22:54 <amaretskiy> each of them will get next RAMInt value
14:23:01 <amaretskiy> same values
14:23:18 <amaretskiy> but they can add own index ro this value
14:23:22 <amaretskiy> for example
14:23:29 <boris-42> amaretskiy: nope
14:23:36 <amaretskiy> we started 3 distributed rally instances
14:23:42 <boris-42> amaretskiy: nope
14:23:45 <amaretskiy> ?
14:23:54 <boris-42> amaretskiy: each of runner will have it's unique index
14:24:15 <boris-42> first runner 1 + i * count_of_runners
14:24:25 <boris-42> second runner 2 + i * count_of_runners
14:24:27 <boris-42> and so on
14:25:21 <amaretskiy> so, for example, these runners starts their 1st iterations
14:25:35 <boris-42> amaretskiy: they will have unique iterations ids
14:26:01 <amaretskiy> not clear
14:26:13 <boris-42> what is not clear?
14:26:25 <boris-42> let's assume 3 runenrs
14:26:42 <boris-42> 1 4 7 10 is for first runner
14:26:50 <boris-42> 2 5 8 11 is for second runner
14:26:59 <boris-42> 3 6 9 12 is for thrid runner
14:27:09 <boris-42> each of runners has unique ids of iterations
14:27:15 <amaretskiy> yes
14:27:27 <boris-42> so what is missing in that patch
14:27:33 <boris-42> is that we are not passing scenario id
14:27:42 <amaretskiy> no
14:27:49 <boris-42> as far as I remember yes
14:28:03 <amaretskiy> all these runners knows about tenants and users
14:28:13 <amaretskiy> they must balance them
14:28:22 <boris-42> amaretskiy: it's not runners work
14:28:27 <boris-42> amaretskiy: it's done in context method
14:28:37 <boris-42> amaretskiy: wich don't accept ID of iteration and it should accept
14:29:21 <amaretskiy> i do not see any problems with this patch
14:29:29 <amaretskiy> related to d-runners
14:29:55 <boris-42> amaretskiy: ITERATION_COUNTER = rutils.RAMInt()
14:29:59 <boris-42> amaretskiy: ^ this doens't work
14:30:25 <boris-42> amaretskiy: and to make it work properly you need to pass ITERATION ID to CONTEXT
14:30:37 <amaretskiy> hmmm, yes
14:30:39 <boris-42> amaretskiy: not to generate
14:30:43 <boris-42> amaretskiy: in each context
14:31:17 <boris-42> it's very easy to fix
14:31:18 <amaretskiy> passing iteration number will even work faster
14:31:20 <boris-42> andreykurilin: amaretskiy ^
14:32:52 <boris-42> so that is only blocker
14:32:56 <boris-42> if somebody can do that
14:32:58 <boris-42> that will be nice
14:33:41 <amaretskiy> I believe vponomaryov will finish this patch
14:33:49 <amaretskiy> I will help him
14:33:49 <boris-42> amaretskiy: he was waiting for me
14:33:55 <boris-42> amaretskiy: to finish that part
14:34:05 <boris-42> amaretskiy: so I can finish that part
14:34:08 <boris-42> I will try today
14:34:12 <boris-42> with passing id
14:34:22 <amaretskiy> cool
14:37:33 <boris-42> any other topic?
14:37:44 <rvasilets> nothing from me
14:38:02 <amaretskiy> nothing from me
14:38:57 <andreykurilin> boris-42: one more
14:39:00 <andreykurilin> from agenda
14:39:02 <boris-42> andreykurilin: ?
14:39:32 <andreykurilin> it is not saved:(
14:39:33 <andreykurilin> ok
14:40:01 <andreykurilin> topic: too many commands to display task results
14:40:29 <boris-42> #topic too many commands to display task results
14:40:32 <boris-42> soo
14:40:35 <boris-42> not too many
14:40:37 <boris-42> task report
14:40:40 <boris-42> task results
14:40:46 <boris-42> and task export in future
14:41:33 <andreykurilin> task detailed
14:41:34 <andreykurilin> :)
14:41:56 <boris-42> so actually we should remove task results in favor of task export
14:42:10 <boris-42> and task detailed should be task report with specific key
14:42:22 <andreykurilin> maybe it is better to remove `task detailed`?
14:43:00 <andreykurilin> and use `task results`, `task results --detailed`, `task results --json` ?
14:43:25 <amaretskiy> results and report differs
14:43:33 <boris-42> andreykurilin: so task detailed seems more like report
14:43:47 <boris-42> andreykurilin: so I would keep 2 commands task report (instead of task detailed)
14:43:49 <amaretskiy> "rally task detailed" can be merged into "rally task report"
14:43:53 <boris-42> and task export instead of task results
14:44:02 <amaretskiy> sounds reasonable
14:45:11 <andreykurilin> ok, let's deprecate `task detailed` in favor of task report
14:45:30 <andreykurilin> and `task results` in favor of `task export`
14:45:48 <andreykurilin> lets reduce number of commands:)
14:45:51 <boris-42> andreykurilin: however we will need to keep it depracted forever
14:45:51 <boris-42> =)
14:45:56 <andreykurilin> why?
14:46:06 <boris-42> andreykurilin: a lot of people is using it
14:46:19 <andreykurilin> boris-42: deprecated warning will fix it
14:47:40 <andreykurilin> boris-42: I raised this topic, because I found attempt to introduce one more command to display results...
14:47:50 <boris-42> andreykurilin: export?
14:47:55 <andreykurilin> error
14:48:00 <andreykurilin> or errors
14:48:04 <boris-42> andreykurilin: oh that is very bad UI
14:48:33 <andreykurilin> and there is no real objective against this command, because we already have separate command for different types of displaying results
14:48:47 <andreykurilin> *separate commands
14:49:29 <boris-42> yep we really should not add more commands
14:51:57 <andreykurilin> ok, it looks like we don't have more topics to discuss
14:52:03 <boris-42> #endmeeting