15:02:09 #startmeeting Mistral 15:02:10 Meeting started Mon Jun 19 15:02:09 2017 UTC and is due to finish in 60 minutes. The chair is rakhmerov. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:02:11 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:02:13 The meeting name has been set to 'mistral' 15:02:59 Hi 15:03:26 Hey 15:03:41 hi 15:03:59 let's wait a minute 15:04:03 hello 15:04:06 hey 15:04:50 hiya 15:05:13 hi apetrich 15:05:20 #topic Review Action items 15:05:22 1. rakhmerov: review https://review.openstack.org/#/c/470996/ 15:05:22 done 15:05:26 o/ 15:05:34 2. rakhmerov, d0ugal: figure out for how long old actions should be deprecated before complete removal 15:05:39 done, in R cycle 15:05:43 bobh: hi, welcome ) 15:05:58 I can't wait for R :-D 15:06:03 #topic Current status, plans 15:06:10 d0ugal: kidding me? :)) 15:06:21 I just like to delete stuff :) 15:06:35 d0ugal: but well, yeah, it will be a good thing to get into that crazy threading stuff again ) 15:06:41 yes, agree 15:06:54 ok, so let's share our statuses and plans 15:07:15 after that we can discuss whatever we want, I don't have any specific agenda, just a couple of questions 15:07:23 I don't have much to share, I have been busy with TripleO tasks. I did post this tiny review, which I need tho' https://review.openstack.org/475404 15:07:44 Working on adding a namespace for mistral workflos 15:07:56 my status: reviewing a lot, fixed one bug with trimming "state_info" (also backported it), working on HA (learhing OpenStack CI a little bit) 15:07:58 Workflows 15:08:03 https://review.openstack.org/#/c/470996/ 15:08:18 Just uploaded a new PS 15:09:03 mgershenzon: cool, thanks 15:09:06 I'm changing mistral to use mistral-lib here https://review.openstack.org/#/c/474895/ but right now I'm dead end that is where is a good place to do serializer registration 15:09:22 d0ugal: please also participate in Michal's spec if you have a chance 15:09:25 status: I am at a point where I could use a pair of eye on my proposed change https://review.openstack.org/#/c/455447/ working out a small kink, 15:09:38 btw that patch is a bit old but I'm holding on changes to uplodad some changes 15:09:45 rakhmerov: I will, I reviewed the previous version but not the new one yet 15:10:05 apetrich: yeah, you mentioned it in our channel. I have a couple of ideas but let me think more, I'll leave a comment with my thoughts 15:10:15 rakhmerov, cheers! 15:10:39 with that done. I have some changes to mistral-lib and then move more stuff there 15:11:02 apetrich: great 15:11:10 toure: what do you mean? You need help? ) 15:11:29 rakhmerov: I think toure just wants reviews. 15:11:34 rakhmerov reviews 15:11:38 ooh, ok 15:11:43 d0ugal, I approved the review you requested 15:11:59 :) my sqlalchemy battles are over I think 15:12:02 the reason I haven't reviewed its latest version yet is because it's still marked as WIP 15:12:03 mgershenzon: thanks! 15:12:14 rakhmerov I removed it 15:12:24 it still has -1 workflow 15:12:24 toure: that's very good ) I'll review it 15:12:33 thanks 15:12:54 toure: yeah, just remove -1 Workflow once you think it's ready 15:13:02 k 15:13:04 that is usually a signal that it's ready for review 15:13:11 cool 15:13:17 yep, awesome 15:13:25 looking forward to reviewing it 15:13:58 yeah I was trying to get it ready for pike-2 but sqlalchemy was messing with me 15:14:01 :) 15:14:10 #action rakhmerov: talk to apetrich and decide where to put initialization of serialization stuff 15:14:21 :) 15:14:23 toure: no problem 15:14:33 yep, so 15:14:53 is there any topics you'd like to discuss? 15:15:20 What do you think about a set of Mistral actions for creating and removing temp files/directories? 15:15:26 d0ugal: I'd like to ask you to check the status of that requirement problem 15:15:49 rakhmerov: rbrady was helping me, because I was busy - but he isn't around today. 15:15:55 so I'm not sure if anything changed 15:16:04 yeah, I see, that's why I'm asking :) 15:16:09 Otherwise I should have time to get back to it tomorrow - I have some ideas for a hacky fix :) 15:16:22 because we haven't heard about it for a few days 15:16:30 yeah 15:16:51 I'll find out and try and move it along faster. 15:16:56 I remember you asked me something (there was some problem with keystone again, right?) but don't remember exactly what it was 15:17:01 some decomposition stuff again 15:17:22 too much stuff going on.. 15:17:33 but we can discuss, if needed 15:17:42 ok 15:17:57 d0ugal: on the new actions, can you elaborate? 15:18:04 #topic Open Discussion 15:18:15 temp files/directories where? 15:18:26 on the executors' hosts? 15:18:49 rakhmerov: The problem was with the mistral config, the bit of code we import accesses the Mistral config - so moving it out gets tricky. 15:19:04 rakhmerov: tempfiles - yeah, it would be on the executors hosts - but that would be a problem in a HA setup 15:19:07 aah, yeah, right 15:19:11 so I am thinking it isn't a good idea now :) 15:19:31 what's your use case? 15:19:36 it is a request I got from a tripleo dev, he needs to create a temp directory to pass to ansible 15:19:49 HA setup you mean when we have many executors on different hosts? 15:20:25 yeah 15:20:31 yes, right 15:20:36 meaning that the temp file might not exist on the host you need it on 15:20:49 however, there's one thing that not many probably use called "Task affinity" 15:21:01 or "task-executor affinity" 15:21:33 it's when you can mark a task with "target" attribute and its actions will always be routed to corresponding a executor 15:21:48 right 15:22:07 still, I think the tempfiles stuff could be a bit strange. 15:22:11 so potentially it might help if used with precaution 15:22:22 yup 15:22:22 true 15:22:35 it sounds a bit messy for me 15:22:41 we've never considered Mistral as a file system aware thing 15:22:45 yes 15:23:13 btw, I now remember that somebody already proposed a similar thing 15:23:16 I guess kong 15:23:26 there still should be a WIP in gerrit 15:23:30 ah 15:23:33 I'll take a look 15:23:37 may be we need to get back to reviewing it 15:23:38 yeah 15:23:59 ooh, it was something like "Run a script" action 15:24:12 Maybe openstack needs an object storage service. Do we have one? 15:24:13 that runs different types of scripts or binaries 15:24:18 depending on how it's configured 15:24:46 mgershenzon: well, we have Swift (S3-like) or Glance-Glare 15:25:12 Glare is a cool stuff which we still need to integrate with 15:25:43 d0ugal: yeah, but that patch got stuck in a long discussion. Lots of concerns were about security, as far as I remember 15:25:49 yeah :) 15:26:12 kong was proposing it more for private cloud use cases 15:26:21 mgershenzon: we might update it to work with Swift - the problem is the code currently expects a temp directory 15:26:25 where there's more control over infrastructure 15:26:32 and for a moment it made sense that we had actions for tempfiles 15:26:39 then I remembered about multiple executors :) 15:26:58 d0ugal: take a look at "target", it may really help 15:27:03 I see 15:27:17 that way we'll always know where a certain action will run 15:27:17 rakhmerov: will do. 15:27:37 however, somebody reported recently that this feature is now broken :) Haha 15:27:55 I don't know if it's true (saw no evidence yet) but we'll need to check 15:28:39 Maybe if we group executors by VM and require tasks to run in a specific group it will work 15:28:47 d0ugal: one more question, do you think you'll have time soon to move OpenStack actions to mistral-extra? 15:29:10 mgershenzon: yes 15:29:34 rakhmerov: I really want to, that task has the same problem that tripleo has. mistral-lib needs access to the mistral config. 15:29:48 so once I have a solution for that, I can move both items forward. 15:29:55 Suggestions are welcome :) 15:30:30 you mean the config problem? 15:31:24 Yes 15:31:33 my solution: it requires a serious refactoring first, at least we need to decompose things properly first 15:31:47 it's now too entangled 15:32:26 #action rakhmerov: look at the config problem (some actions need Mistral config and can't be easily extracted) 15:32:47 ok, I'll have to look at it too 15:32:58 (gosh, so many things..) 15:32:59 :) 15:33:33 ok 15:33:44 is there anything else? 15:34:16 I'd also like to have a bug triaging session, maybe next meeting? 15:34:31 I believe we can close some of the bugs 15:35:03 I think that is all from me 15:35:09 A bug triage would be good. 15:35:32 ok 15:35:50 so let's end the meeting then? 15:35:58 thanks for coming 15:36:05 have a good week ) 15:36:05 Fine by me 15:36:09 bye 15:36:12 Bye 15:36:12 have a good one 15:36:17 #endmeeting