17:05:24 #startmeeting Rally 17:05:25 Meeting started Tue Aug 19 17:05:24 2014 UTC and is due to finish in 60 minutes. The chair is boris-42. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:05:27 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 17:05:29 The meeting name has been set to 'rally' 17:05:47 k4n0 rediskin ping 17:07:11 hello 17:08:50 Hi all 17:08:59 hi 17:09:14 Hey 17:09:32 Hi 17:10:38 Okay let's start 17:11:19 #topic Making a small virtual summit 17:11:35 guys what do you think if we make in hangout mini summit? 17:11:50 e.g. 2 days where we can discuss interesting stuff 17:13:03 boris-42, +1, when are we planning this, I am okay with sometime around mid-sept 17:13:23 coolsvap probably mid-august? 17:13:38 coolsvap mid september will be quite close to openstack summit 17:14:39 boris-42, its already Aug 19 maybe around second week of sept 9,10 17:15:00 coolsvap oh shii 17:15:02 =) 17:15:22 coolsvap okay 9,10 september seems ok 17:15:43 coolsvap we wil have time to prepare docs for discussion 17:15:52 boris-42, yup 17:16:10 olkonami temujin rook what do you think ? 17:16:41 sept 9,10 is ok for me 17:16:49 is this a virtual event? 17:17:07 rook yep 17:17:18 rook just a hangout that can be join by everybody 17:17:34 boris-42 I am good with the September dates. 17:18:33 marcoemorais harlowja ^ 17:18:44 think, that's all right for such event 17:18:45 boris-42 cool 17:19:19 boris-42 coasterz i think i'm fine with sept 9,10 17:19:26 oops, should be coolsvap 17:19:37 harlowja ok great 17:19:49 it will be great opportunity to collect user expireicne 17:19:55 and to dicuss how to cover more use cases 17:20:02 okay let's move to next topic 17:20:16 #topic "rally info" command 17:20:35 #link https://review.openstack.org/#/c/102853/ 17:20:42 ^ This is the start of long road 17:21:04 where you will be able by running CLI commands to get all interesting infromation 17:21:13 so this is first step 17:21:31 "rally info find NovaServers" will print it's docstring 17:21:48 as well I think that in future it should print available benchmarks 17:22:22 ally info find NovaServers.boot_and_delete_server will print detailed info about benchmark and what arguments it accepts (if this info is presented in code) 17:22:42 so we should improve quality of some doc strings 17:22:49 and work on usability of this comamnd 17:23:32 e.g. listing stuff, and find that supports regexp and so on 17:24:12 rook coolsvap k4n0 harlowja olkonami temujin ^ 17:24:18 any thoughts? 17:24:44 boris-42, agreed we need to work on doc strings if we want to increase its usability 17:25:10 coolsvap yep so probably we need to add blueprint 17:25:15 coolsvap like we have for tests 17:25:29 #action add blueprint for improving doc strings in benchmarks 17:25:33 boris-42, yup 17:27:42 okay let's move to next topic 17:27:57 #topic generic cleanup 17:28:04 Okay we are still fighting 17:28:15 to get clean up work in every case in any case 17:28:32 so to do that rediskin is refactoring current cleanup context 17:28:37 rediskin are you here? 17:30:09 okay I'll try to cover most important points 17:30:41 Split admin & user (cleanup) already done. This is required case we are going to support benchmarking withou admin access 17:30:56 cause* 17:31:10 And it's already done 17:31:31 The second thing is to create and extensible framework for adding new resources to cleanup 17:31:46 So users will just specify some class and implement 1-2 methods 17:31:53 that cleanup single resource 17:32:18 and cleanup context will know how to use these mini-resource cleanup's 17:32:43 and do all logic with limiting speed of resource deletion 17:32:48 and repeat in case of failuers 17:33:07 https://review.openstack.org/#/c/113536/ 17:33:19 there is already patch that remove part of crap ^ 17:33:49 And as well there is third point that should be done 17:33:51 it will be quite big 17:34:08 it's finish unification of resource names & implement command 17:34:18 that will delete all resources with this specifc pattern 17:34:34 e.g. rally__ will be the name of resource 17:34:43 so we can list all resources and delete one by one 17:35:04 (in any case even if something went wrong with server that is running rally) 17:35:25 coolsvap rook harlowja olkonami temujin ^ 17:35:52 i'm still a fan of writing a file that has all the things rally created ;) 17:36:02 or multiple files... 17:36:14 harlowja it will create a big overhead imho 17:36:23 meh 17:36:30 harlowja plus what if something happen with that file 17:36:36 like aliens take it? 17:36:41 or comets hit it? 17:36:44 that's grate idea to rollback changes the benchmark made 17:36:45 space aliens 17:37:27 harlowja yep 17:37:30 I prefer approach from boris-42 17:37:38 ok dokie, up to u guys 17:37:46 i like AOF for this kind of stuff :-P 17:38:01 harlowja making file in case of distributed load 17:38:02 *append only files 17:38:05 harlowja maybe dangours=) 17:38:08 meh 17:38:17 life is dangerous 17:38:21 i drove my car in today, that was dangerous 17:38:22 lol 17:39:29 harlowja lol 17:39:31 harlowja yep 17:39:33 :) 17:39:45 but u know what, i did it! 17:39:48 amzing, i know 17:39:52 harlowja I really don't know we can provide this way with creating resources =) 17:40:04 harlowja and putting them to file 17:40:15 harlowja but really not sure that it's better then just don't have file=) 17:40:26 easier to view when just a bunch of files 17:40:40 easier to make non-monolothic cleanup scripts 17:40:40 :-P 17:40:48 i heard rally is monolothic, lol 17:40:53 harlowja me too 17:40:59 harlowja too monolothic 17:41:02 :) 17:41:15 lol 17:41:24 it is just a bunch of scripts 17:41:31 rook rally? 17:41:34 monolothic godamnit 17:42:00 rook actually it's in one directory so it's monolithic.. 17:42:12 omg 17:42:14 crazy 17:42:15 lol 17:42:28 I spent most of my lunch reading through that thread 17:42:34 rook ahaha=) 17:42:42 rook I eat already tons of popcorn=) 17:43:43 I don't have a dog in the fight... Just interesting to read the different opinions 17:44:41 rook so i really dislike of splitting rally to separated repositories in the way that QA wants 17:44:49 it will make development and usage harder 17:44:58 well I heard it is unstable and will break your cloud and kill rook's dog 17:45:24 lordd_ you assholes! that is why my dogs keep dying! 17:45:40 :D 17:45:43 boris-42 can you explain what this split is? 17:45:43 lol 17:46:01 rook so they would like to move in separated repository load generators from rally 17:46:08 rook scenarios as well in separated 17:46:21 rook and script that save data to DB in separeted 17:46:39 rook and stuff that generates html with results in separate 17:46:44 is there then a script that combines all these back together to form rally, lol 17:46:55 harlowja yep 17:46:58 harlowja in gates 17:47:16 wtf 17:47:22 is Tempest handled in this way? 17:47:22 rook so and all this to be able to consume these scripts with tempest 17:47:32 rook they would like 17:47:33 thats like captain planet right, where u combine all the rings and stuff 17:47:43 harlowja: hahaha 17:47:55 harlowja lol 17:48:03 lol indeed 17:48:18 so I really don't know how to convince people 17:48:22 just to left rally as is 17:48:37 https://www.youtube.com/watch?v=kbW5sxyu9bU (captain planet, for those who have no idea wtf i am referring to) 17:49:26 harlowja ahaha 17:49:34 well, I don't want to be unpopular, but I would like to have an installation method similar to other openstack projects 17:49:54 but I suppose this will come with time 17:49:57 instead of captain planet, it can be captain boris 17:50:11 in any case, with Ubuntu, the install script works perfectly 17:50:42 lordd_ installation for what? 17:50:50 lordd_ you mean be able to do pip install rally? 17:51:00 lordd_ someday someday we will make first rally version 17:51:24 either pip, or OS packaging 17:51:42 but I totally understand why this isn't happening now, and shouldn't be a priority at all 17:51:43 lordd_ so yep someday 17:52:23 this rant was only related to the monolithic, scripting, etc.. 17:52:40 somebody mentioned because of the installation method 17:52:50 lol 17:52:53 in any case, please ignore this comments and continue 17:53:10 lordd_ will de if you say=) 17:53:12 do* 17:53:21 #topic open discussion 17:53:31 okay probably somebody want to put some input? 17:53:55 I have a question concerning network testing 17:54:37 I saw a script example mentioning neutron networks, but that didn't work 17:54:52 lordd_ hm what one? 17:55:08 sorry, the idea would be to associate nova vms to neutron networks 17:55:10 this: 17:56:21 "neutron_network": { 17:56:21 "network_cidr": "10.%s.0.0/16", 17:56:36 this in theory is context for boot_and_delete_server 17:56:41 from NovaServers 17:56:52 context? 17:57:01 "context": { 17:57:07 ah 17:57:15 heh there are 2 patches related to this 17:57:35 https://review.openstack.org/#/c/103306/ 17:58:03 thanks, I will check on that 17:58:12 lordd_ it hold work actually 17:58:16 lordd_ not sure I didn't test it 17:58:19 lordd_ but I'll 17:58:54 ok, next one is to create actual load from benchmarking 17:59:14 lordd_ lets' move to rally chat 17:59:22 ok, np 17:59:24 cause we need to finish meeting 17:59:30 #endmeeting