20:00:58 #startmeeting state-management 20:00:59 Meeting started Thu Dec 5 20:00:58 2013 UTC and is due to finish in 60 minutes. The chair is harlowja. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:01:00 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:01:02 The meeting name has been set to 'state_management' 20:01:06 howdy folks! 20:01:11 hi there 20:01:14 hi 20:01:20 hi there 20:01:51 sooo i guess lets get started 20:02:00 #link https://wiki.openstack.org/wiki/Meetings/StateManagement#Agenda_for_next_meeting 20:02:15 #topic action-items 20:02:19 no action items, haha 20:02:32 #link http://eavesdrop.openstack.org/meetings/state_management/2013/state_management.2013-11-21-20.13.html :) 20:03:14 been 2 weeks, so let me know if i missed anything :), don't think so 20:03:18 #topic current integration 20:03:30 so this is a bigger topic :) 20:03:57 i think cinder is going through, thx anastasia and iv_m and folks :) 20:04:14 #link https://review.openstack.org/#/c/58724/ 20:04:27 that should let others get started on there reviews, victor (and others) 20:04:39 and i think anastasia has a couple followup reviews also 20:05:09 I filed blueprint because I couldn't find same topic. 20:05:17 i think we are all waiting to get that merged 20:05:24 hmmm, what was the blueprint for haruka_ ? 20:05:44 #link https://blueprints.launchpad.net/cinder/+spec/create-snapshot-task-flow 20:05:44 #link https://etherpad.openstack.org/p/TaskflowIcehouseWhoWhatWhereWhy (is where i've been connecting everything together) - if u want to put a link there 20:06:09 k, thx haruka_ added to etherpad 20:06:17 thank you 20:06:37 np, launchpad doesn't have such a good way to connect everything as releated to taskflow 20:06:50 so that etherpad has been serving this purpose 20:07:09 ok 20:07:23 iv_m i'll see if i can bug the cinder folks for that review later 20:07:34 great 20:07:38 see if we can get them to approve it 20:07:43 since its got the +2 20:08:38 o/ 20:08:40 #link https://review.openstack.org/#/c/59327/ is also being worked on by the mistral folks 20:08:51 howdy MikeSpreitzer 20:09:18 and lots of other blueprints :) 20:09:56 iv_m i think u are working through https://blueprints.launchpad.net/nova/+spec/glance-snapshot-tasks (or part of it?) 20:10:04 yup 20:10:13 you're typing faster than me) 20:10:17 ;) 20:11:02 cool, haruka_ the steps in above bp glance-snapshot-tasks seem like they would be applicable for your create-snapshot-task-flow (and others also) 20:11:02 and aslo Alexander Gorodnev is helping me there 20:11:08 great 20:11:18 now we're still at #0 of that bp -- understanding and documenting 20:11:34 nikhil__ and markwash maybe can help ensure sanity there ;) 20:11:57 iv_m totally understood, #0 is a big part of it 20:12:05 oh snap, that reminds me to take a look at the bp :) 20:12:09 :) 20:12:20 most of actual snapshoting workflow happens inside single nova's driver method -- so looks like we'll have to adjust driver api 20:12:33 ya, the blackbox problem 20:12:45 hard to make a blackbox reliable 20:13:13 #link http://en.wikipedia.org/wiki/Black_box (for those who have no idea what i am talking about, haha) 20:13:32 nikhil__ if u get some time that would be super appreciated 20:14:07 sure harlowja (I've the link open in my browser, just need to visit it) 20:14:21 #action iv_m agorodnev document nova's snapshoting flow and what we want to do with it 20:14:24 i might see if boris-42 wants to take on https://blueprints.launchpad.net/nova/+spec/terminate-instance-tasks, if not i'll do that, gotta catch up with him on that 20:14:41 he's pretty familar with nova (so thats good) 20:15:05 #action harlowja catch up with boris-42 on https://blueprints.launchpad.net/nova/+spec/terminate-instance-tasks and if he wants to own it (or should i just own it) 20:15:49 the other new one for this week is https://etherpad.openstack.org/p/Rally_TaskFlow_Integration 20:16:02 also working with boris-42 there to see how it might fit 20:16:16 feel free to add any comments there :) 20:16:28 or #openstack-rally to 20:17:08 nikhil__ for glance itself, are u (or other?) still working on the async stuff 20:17:38 i think with the work @ https://review.openstack.org/#/c/59144/ (ongoing) taskflow gets closer to what glance desires 20:17:46 harlowja: yeah, we'r actively working on it 20:17:59 3 patches went in and bunch are to follow 20:18:03 cool 20:18:35 https://blueprints.launchpad.net/glance/+spec/async-glance-workers still is the 'root' blueprint for all those right? 20:18:48 the dependency tree on that one really looks like a tree :) 20:19:02 yeah, that sounds right 20:19:03 loopsided christmas tree, ha 20:19:06 that looks scary 20:19:09 :) 20:19:11 I prefer to not look at it 20:19:21 :) 20:19:59 k, np, i'll look at whats in glance, and maybe write up some ideas for taskflow (or in the future how said glance concepts might translate to taskflow) 20:20:16 thx nikhil__ ! 20:20:50 harlowja: sounds good 20:21:15 :) 20:21:22 will keep looking into that MP 20:21:29 thx much 20:21:49 and current integration blockers that we can help resolve that people are facing (the cinder 0.1.1 should unblock folks) 20:22:05 *any current 20:22:54 if not, thats ok to (even better, ha) 20:24:07 #topic taskflow features 20:24:34 just a little overview of what i know about being worked on in taskflow (others feel free to add) 20:24:46 1) zookeeper work (by me) 20:24:59 i was thinking i can make a little demo of this for folks, since it might not be obvious how neat it is :) 20:25:24 would be great 20:25:28 maybe some little tutorial or something 20:25:31 with some code 20:25:51 cool, iv_m i'll see what i can do :) 20:26:04 something that others can pretty easily also try 20:26:16 i was thinking to give it really serious and hard review soon, so being able to see it in action and play with it would help 20:26:21 agreed 20:26:36 #action harlowja make little tutorial/example (other) for zookeeper review 20:27:34 2) the oslo.messaging worker distributed neato stuff - https://etherpad.openstack.org/p/TaskFlowWorkerBasedEngine (and review WIP @ https://review.openstack.org/#/c/59144/) 20:27:50 ^ above i think is waiting a little bit on oslo.messaging being a little more library like 20:28:17 we talked with Stanislav (who did most of that work) and decided to put it off a bit 20:28:26 because of oslo.messaging 20:28:45 ya, understandable 20:28:59 now we're looking on what interesting things we can do using kombu directly 20:29:12 kk, #link https://review.openstack.org/#/c/59728/ (and a few others) are underway 20:29:26 to try to make it more python3 compat, and oslo.cfg independent... 20:30:13 #link https://review.openstack.org/#/c/59758/ (for some more python3 stuff) 20:31:07 looking forward to see what that becomes (its a tough challenge) 20:31:46 iv_m for the reversion strategy stuff (which i think u are working on) is there any recommended link u have for that 20:32:16 not yet, sorry 20:32:27 np 20:32:32 all good 20:33:34 so one of the other ideas that happened yesterday (still being baked, not concrete) is https://etherpad.openstack.org/p/async-taskflow-tasks 20:33:53 if people want to add any thoughts, questions, comments on there that'd be great 20:34:25 hopefully the use-case is clear (the ideas on how to solve might not be, if not, feel free to speak up) 20:34:55 so just throwing that out there (feel free to check it out) 20:35:24 ok. I will check it. 20:35:29 sounds great 20:36:13 any taskflow features that need any discussion, i think changbl is working on zookeeper stuff as well 20:36:31 #link https://blueprints.launchpad.net/taskflow/+spec/zk-logbook 20:37:31 yes, working on it 20:37:35 sweet, thx changbl :) 20:37:49 let me know if u get stuck, willing to help out wherever needed! 20:37:58 sure 20:38:52 just a quick annonce -- me and Anastasia started work on refactoring engine internals, which is needed for worker-based engine, help in reversion strategies and may bring support of multiprocessing executors 20:39:15 too early to say anything, maybe next time we'll have something cool to show) 20:39:27 alright, lets see what happens there ;) 20:39:42 just internals i hope :) 20:40:35 ya, we are not intending to break anything (like api) right now 20:40:41 sweet 20:41:51 thx iv_m , eagerly awaits next week now, haha 20:42:43 ) 20:43:02 any other new taskflow features/goodies that people want to talk about? 20:43:10 can just switch to open-dicuss for anything else 20:43:54 3 20:43:59 2 20:44:04 1 20:44:11 #topic open-discuss 20:44:45 iv_m will your internal engine work have ideas from https://etherpad.openstack.org/p/async-taskflow-tasks (out of curosity) 20:45:49 if its to early to tell, thats fine 20:46:03 no, it started a bit earlier 20:46:18 np 20:46:44 http://www.python.org/dev/peps/pep-3156/ is an interesting read for folks also (the future of python concurrency in a way) 20:47:07 *or maybe the future (still hard to tell, ha) 20:47:14 yeah, it's already in python 3.4 alpha 20:47:43 lots of new concepts there, but a little overlap with taskflow (in a way) 20:47:51 * iv_m lost count of 'future of the futures' puns 20:47:56 :) 20:48:25 ya, i should grep for how many mentions that page has of 'future' 20:48:29 probably over 100 20:49:13 they have most of interface in common with concurrent.futures, so i think if we're nice with that we'll be prepared) 20:49:19 def 20:50:00 be interesting to see what happens in this area (especially with releation to openstack) 20:50:31 and how this works with eventlet (and eventlet+openstack...) 20:50:38 fun times ahead, ha 20:51:29 anyways, can chat more about this in #openstack-state-management so as not to keep people around if there isn't anything so useful to discuss :) 20:51:59 since i know iv_m probably wants to go to sleep, haha 20:52:44 other news, i think boris (from mirantis) will start getting some taskflow programemrs involved from his side, so +1 (+2 when i see the people, ha) 20:52:54 u know my weeknesses 20:53:19 :) 20:53:34 k, letting people go, 7 minutes left, ha 20:53:44 i filed one more bp #link https://blueprints.launchpad.net/nova/+spec/feature-of-cancel with using taskflow 20:53:46 ah 20:53:55 just under considering 20:53:58 intersting 20:54:10 cool, haruka_ will check it out 20:54:18 and will add some comments 20:54:23 much appreciated! 20:54:41 ok 20:55:30 just did a quick look over, seems to be the right kind of idea 20:56:01 i'd use that for like a top level blueprint and add in dependents for the pieces 20:56:23 that seems to be what russellb wants for that one (in whiteboard) 20:56:36 yes. 20:57:19 shall i help u there, or do u have that covered, can see if we can brainstorm how to make that possible 20:57:33 splitting the work and figuring out the pieces is not always easy :) 20:58:51 k, running out of time 20:59:10 haruka_ if u want to brainstorm on how to do this we're always in #openstack-state-management channel (most of us anyway) 20:59:11 yes. thank you. Your comment is always appreciated. 20:59:16 great 20:59:26 as for others, #openstack-state-management also :) 20:59:33 until next time! 20:59:42 thx for coming folks :) 20:59:56 #endmeeting