20:00:55 #startmeeting state-management 20:00:56 Meeting started Thu Dec 12 20:00:55 2013 UTC and is due to finish in 60 minutes. The chair is harlowja. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:00:58 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:01:00 The meeting name has been set to 'state_management' 20:01:02 hi folks! 20:01:07 hi there 20:01:09 hi hi 20:01:11 hi 20:01:17 just got out of sprint meeting, let me update agenda 20:01:56 *internal y! sprint stuff... 20:02:01 hi 20:02:44 hi mspreitz iccha k4n0 20:02:51 ok, updated 20:02:53 #link https://wiki.openstack.org/wiki/Meetings/StateManagement#Agenda_for_next_meeting 20:03:21 guess lets do a quick last-action-items overview 20:03:27 #topic action-items 20:03:33 #link http://eavesdrop.openstack.org/meetings/state_management/2013/state_management.2013-12-05-20.00.html 20:03:49 so i haven't caught up with boris yet (will try to do today) 20:03:58 as for little zookeeper example (to help out with review) 20:04:22 paste.openstack.org/show/54861/ is my current incarniation of that 20:04:30 steps at top should help 20:04:57 #action harlowja really catchup with boris this time! 20:05:25 if u want i can try to catch him in skype 20:05:39 i can usually catch him on IRC, just i totally forget to ask him :-) 20:05:52 if i don't iv_m then lets skype 20:06:02 *hopefully today, if he's still on 20:06:05 or tommorow 20:06:16 thx iv_m 20:06:35 iv_m any update on the snapshot nova flow and work? 20:07:12 so, i failed my action item -- we are working on it, and came to some conclusion, and started to code it -- but no docs yet, sorry 20:07:31 np iv_m :) 20:07:44 still being worked on, thats fine with me :) 20:08:09 * iv_m just needs to write everything down 20:08:12 :) 20:08:18 #action iv_m write all the things down! 20:08:18 ha 20:08:24 all the things please :-P 20:08:36 ha 20:09:01 well, that'll need a lot of space down there 20:09:15 :) 20:09:18 and alot of time 20:09:28 but i'll try to do my best anyway 20:09:41 haha, sweet 20:09:48 #topic current integration work 20:10:01 so i think we covered one of the current integration work that is ongoing (nova) 20:10:40 akarpinska1 is nearly there with the basic cinder work (thx akarpinska1 !) 20:11:10 although i think that one might be hitting some other major gate bugs 20:11:13 #link https://review.openstack.org/#/c/58724/ 20:11:24 no so easy to tell 20:12:05 ya, looks like figting through gating bacame harder lately 20:12:10 ya 20:12:13 #link http://lists.openstack.org/pipermail/openstack-dev/2013-December/022052.html 20:12:16 is it necessary to do rechecks until I get +1 from Jenkins? 20:12:32 yup 20:12:45 akarpinska1 if there is an active gate bug that is being fixed, then u can sorta wait till its fixed 20:13:01 but ya, in general, jenkins needs to +1 it (which means the gate must not be broke) 20:13:15 the #openstack-infra team might have an idea whats going on (seems to change daily) 20:13:48 akarpinska1 so close though! :) 20:14:11 i think other folks that are getting involved with taskflow + cinder very much appreciate your work :) 20:14:39 anyone that is also working with cinder & taskflow want to add any questions/other (think there are others out there somewhere) 20:15:28 ok, well don't be shy folks :) 20:15:45 * akarpinska1 doing recheck 20:15:54 thx akarpinska1 20:16:07 as for other integration work, i think boris-42 is still experimenting with rally + taskflow 20:16:23 and same with mistral project 20:16:54 as usual 20:16:56 #link https://etherpad.openstack.org/p/TaskflowIcehouseWhoWhatWhereWhy 20:17:04 *where i track most of this 20:17:58 alright, keep up the integration work folks, slowly but surely :) 20:18:05 #topic new-use-cases 20:18:52 so any new issues or use-cases people want to talk about for taskflow 20:19:12 i started a little document that helps explain some of the mind-set change that taskflow might make happen 20:19:25 releated slightly to use-cases 20:19:26 #link https://wiki.openstack.org/wiki/TaskFlow/Best_practices#Mind-Altering_Substances 20:19:32 *no real substances there 20:19:55 nice little place to gather common ideas, patterns, and such 20:20:57 as for other new-uses, iv_m and akarpinska1 and a few others have been continuing to work on the 'worker' engine model 20:21:03 which should enable some more neat use-cases 20:21:38 also i expect https://review.openstack.org/#/c/60499/ (the tooz project) when combined with taskflow will also add even more 20:21:46 #link https://github.com/stackforge/tooz 20:22:06 anyways, lots of room for new usage :) 20:22:25 #action harlowja describe jobboard mindset change on best practices page 20:23:54 cool, anyways, feel free to bring new usecases and patterns :) 20:24:20 always will to discuss them with folks 20:24:27 #topic open-discuss 20:24:41 iv_m do u want to give a little overview of the changes in the executor (futures) that u are thinkign about? 20:24:55 * harlowja needs to get around to reviewing the changes that u did 20:25:07 or maybe still to early for said discussion (which is fine) 20:26:17 well, the general idea is to have special entity that will know how to take task, its arguments, and apply one to another 20:26:55 it should have closer scope than current TaskAction, wich knows about storage and manages task state changes and suchlike 20:27:17 sure, makes sense, to more focus just on executing, not the rest of the stuff 20:27:45 so, #link https://blueprints.launchpad.net/taskflow/+spec/task-executor 20:27:57 ah, thx for sharing 20:28:04 good description there 20:29:04 next step is to allow task executor to run task asynchronously 20:29:15 :) 20:29:24 that's what worker-based engine will need to do, for example 20:30:01 right 20:30:11 so, the executor should return future that will in order return task result (or task reversion result) and some metadata 20:30:39 cool 20:30:53 what are u thinking about for the metadata? 20:31:24 well, in my experimental code it's task instance itself and what happend (was it reverted or executed) 20:32:29 something like that so that engine does not have to have some sophisitacted data structure to find out where result returned by future belongs 20:32:29 k 20:33:02 in experimental code, i guess listeners that are watching the state transitions still operate as before 20:33:27 of course 20:34:21 neats 20:34:59 then, the plan is to create new graph action that will be able to schedule all the tasks with task executor and wait on returned futures while tasks are executed (maybe in parallel), using no more than one thread 20:35:59 I'm not shure it is possible to do in one thread 20:36:11 because of feature implementation 20:36:16 hey, i almost already done that 20:36:38 do you implement your own feature? 20:37:07 no, i use concurrent.futures.Future in my code 20:37:44 but it waits until result is set 20:37:44 interesting 20:38:10 u can wait for several futures at once with concurrent.futures.wait() 20:39:25 wich makes everything really simple 20:39:44 something though blocks right? so for something to block u need something to block on (another thread?) - so that means 2 threads? 20:39:53 anyways, techinical detail, lol 20:40:09 *this might of been what akarpinska1 was asking (not sure) 20:41:39 oh ya, other thing i was going to bring up 20:41:41 #link https://etherpad.openstack.org/p/async-taskflow-tasks 20:42:07 i think with iv_m work here, #3 becomes more natural (although #0 still seems like it would also be fine) 20:42:22 so i can try out #3 when iv_m executor stuff in place 20:42:26 see how that goes 20:42:39 yup 20:42:43 might not be so bad, not so sure yet 20:43:12 that will make mistral folks use-case work better (maybe?) 20:44:02 what else, oh ya, nearly forgot, changbl i think wasn't able to make it today, but he said offline that he is working on the zookeeper stuff (ongoing) 20:44:07 and getting closer! 20:44:24 will keep all of us posted 20:44:29 *changbl will 20:45:04 great 20:45:43 :) 20:46:09 * harlowja needs to track down some more of the cinder folks doing taskflow integration, see where that is at 20:46:34 #action harlowja see if can track them any of those folks down (to see if they are doing ok) 20:46:52 i know they exist somewhere :-P 20:47:12 ) 20:47:56 will see if i can fire some folks an email that might know 20:48:28 anything else people want to talk about, i think i'm all good 20:49:15 goiinnng once 20:49:56 goinnng twice 20:50:15 going thriceeee 20:50:23 sold to iv_m for 500$ 20:50:26 ha 20:50:50 np, but what did i get? 20:50:52 ha 20:50:57 secret prize 20:51:34 anyways #openstack-state-management for any further questions, issues, problems, q/a, chit-chat, prizes... 20:51:54 until next time folks, and thanks for coming for those that did :) 20:52:00 thanks guys 20:52:03 np :) 20:52:22 #endmeeting