20:01:20 #startmeeting Octavia 20:01:20 Meeting started Wed Jun 10 20:01:20 2015 UTC and is due to finish in 60 minutes. The chair is xgerman. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:01:21 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:01:24 The meeting name has been set to 'octavia' 20:01:27 o/ 20:01:27 o/ 20:01:29 o/ 20:01:35 #chair blogan, dougwig 20:01:36 Current chairs: blogan dougwig xgerman 20:01:48 (I am in the lunch room not sure when they kick me out) 20:01:50 o/ 20:01:50 o/ 20:02:02 hi again 20:02:07 #topic Announcements 20:02:18 Midcycle: https://etherpad.openstack.org/p/LBaaS-FWaaS-VPNaaS_Summer_Midcycle_meetup 20:03:30 Protected instances- https://openstack.nimeyo.com/47616/openstack-neutron-octavia-protected-openstack-resources 20:03:43 o/ 20:03:56 it seems the nova guys are very self assured of their abilities ;-) 20:04:16 Lastly, Global Load Balancing - https://docs.google.com/document/d/1016shVnPaK8l8HxMpjADiYtY2O94p4g7JxjuIA-qv7w/edit 20:04:28 anyhting else? 20:05:35 sounds like a resounding no 20:05:49 I have been under a rock the last three days) 20:05:57 #topic Brief progress reports 20:05:57 have you? 20:06:22 I have been filling in for folks for the last few days and feel like I was under a rock 20:06:26 well, internal conference — today is the first day I figured out how to use the Internet 20:06:28 still working on internal stuff, but let me know if there are reviews that need eyes 20:06:53 I need reviews on 3 reviews that I'll put in links in a second 20:06:58 kk 20:06:58 cool 20:06:59 #link https://review.openstack.org/#/c/170989/ 20:07:01 I have sonar and jenkins installed for our sonar gate pilot. I need to work on getting the plugins setup 20:07:07 #link https://review.openstack.org/#/c/187237/ 20:07:14 #link https://review.openstack.org/#/c/190269/ 20:07:20 johnsom +1 20:07:59 #link https://review.openstack.org/#/c/187658/ 20:08:26 Could use some eyes. The first fix solved the taskflow issue, but uncovered a model issue 20:08:50 If someone has spare cycles feel free to mess with it. 20:09:21 johnsom: if its a data model issue i might take a look 20:09:49 My quick test showed it was erroring while deleting the member from the model 20:10:18 mmh 20:12:02 ill take a look, ive got the data model changes for the networking and amphora drivers almost completed 20:12:07 seems like al ogical next step 20:12:12 guess moving on... 20:12:13 #topic Open Discussion 20:12:31 ive got one 20:12:36 shoot 20:13:07 so the flows we have right now assume that the amphora can be hot plugged 20:13:18 we, rax, are going to be using containers 20:13:33 we might, too 20:13:35 and the current flows will not work in this case so we'll need to create some alternate flows 20:13:44 what do yall think is the best way to do that 20:13:57 Yeah, that is interesting 20:14:09 do if statements in the controller worker, implement a new controller worker based on containers, etc 20:14:19 I was thinking in the beginning to make the flows configurable in the config file 20:14:23 So, basically we have to define all of the plugging at creation time? 20:14:38 xgerman +1 20:14:42 johnsom: yep, so that might mean not creating the amphora until the first member is created 20:14:55 xgerman: explain that in more detail 20:15:08 xgerman, that sounded like "flow drivers" 20:15:08 like have a hot_pluggable config option? 20:15:24 and we just do if statements based on that value? 20:15:34 well, you could say: Create_Amphora_flow: and then list the tasks 20:15:43 and the system would read and build that flow that way 20:15:51 at the startup time 20:16:00 and you do that for all flows 20:16:17 that allows you to have different tasks for different operators 20:16:35 Yeah, that was a longer term goal 20:16:54 yeah that sounds very ambitious right now 20:17:04 good idea though 20:17:41 Is plugging on the roadmap for the container tech you are planning to use? LX*? 20:18:03 its on a very distant roadmap and im not sure if once we get around it if it will ever be done 20:18:10 but if magnum gets it in then we might 20:18:19 but magnum doesn't have any immediate plans eitehr 20:18:33 from a quick conversation with adrian 20:19:00 Hmm, so limits members to a single network 20:19:22 not necessarily but it does make it more complicated 20:19:23 not necessarily, you might be able to plug multiple networks at creation 20:19:30 yeah you can do that 20:19:42 but then you have to worry about killing the machine and respinning it everytime 20:19:57 a new member on a new network is added 20:19:58 or task flow does ;-) 20:20:05 I would prefer we get creative, for example configurable flows, than putting a bunch of 'if' statements in. 20:20:09 (also why a single create call is so valuable, we get the networking configuration up front) 20:20:41 I don’t think configurable flows are that tough to do 20:20:45 but I am also naive 20:20:58 yeah i dont like the if statements either, and the controller worker already adheres to interface, which is why implementing a container controller worker was an option 20:21:30 I am warming to the configure it up front and kill/rebuild for plugs. If we have active/standby that could be transparent. 20:22:21 that might be a way we can go ahead right now and then fill in the other stuff later as an optimization? 20:22:40 (aka if you can hot plug use that flow) 20:23:28 but it feels we have three options and we dislike one (ifs) 20:23:47 so do we need to do a “spike” - or can we make some decision right away? 20:23:53 unless ifs are super simple right now and not ugly :) 20:24:35 Let's think about and research loading flows. Do we have the flows defined in a way that we can swap those two use cases in and out, etc? 20:24:51 +1 20:25:03 we would need to figure out if we need additional tasks anyway 20:25:11 or change tasks 20:25:34 yeah thats a good research task 20:25:55 ill probably research taht too bc i'd like to get something like that in asap bc we've got a lab all set up ready to go with what we need 20:26:03 just need to get the code in there to handle containers 20:26:10 cool!! 20:26:23 do you think you can share performance numbers? 20:27:39 i don't see why not, but id probably have to ask someone first 20:28:45 it is getting very interesting rihgt now though and id love to share everything 20:28:48 Nope. Sharing isn't supported in the Openstack Community. RIP 20:28:52 lulz jp jp, we can check 20:29:35 thanks — I am hoping we can do some work with vms to get some numbers, too 20:29:57 it is nice to have a container immediately 20:30:02 Yeah, and try out the various tech, lxd, docker, vms, etc. 20:30:13 nova boot, nova list, its already ACTIVE 20:30:28 neat 20:32:20 im sure the first pass at doing this for containers will change drastically over time 20:32:29 same with VMs as well 20:32:42 yep 20:33:19 alright anyone have anything else? 20:33:25 Yeah, and whatever else comes around 20:33:30 When is the next mid-cycle again? 20:33:38 xgerman linked it above 20:33:46 and its on the agenda wiki 20:33:50 yeah, scroll to the top 20:34:07 7/15 - 7/17 20:34:30 but make sure to book soon - hotels are real tight - otherwise you will live in Tukwila 20:34:43 or a dog house 20:34:53 yep, we can ask Hank 20:35:06 Its okay, we'll party-crash dougwig's room 20:35:13 (Hank outs dog pictures in all his powerpoint) 20:35:30 hank the hound dog 20:36:02 barclac offered his yard 20:36:14 ill take his plane 20:36:51 blogan you haven't seen the size of his plane have you... 20:37:08 but he doesn’t have a dog house 20:37:08 sleep in the hangar 20:37:21 You could sleep on the wing I guess 20:37:24 would it be like an elephant in a dog house? 20:37:53 okay i think we've digressed enough :) 20:37:53 Only other topic to raise, dougwig mentioned the octavia rename is Friday. 20:38:16 yeah, I think I volunteered to watch and eat popcorn 20:38:22 Me too 20:38:22 Who actually handles the renaming? 20:38:26 infra 20:38:35 i gave anteaya your names, xgerman and johnsom 20:38:40 So they'll update devstack with appropriate endpoints and such as well or no? 20:39:02 most importantly they will invite me and johnsom to the core party 20:39:06 TrevorV: no. they handle the git rename, gerrit patch updates and such. 20:39:23 anywhere we've hardcoded a git url is on us to fix. 20:39:55 ok, probably in our devstack config samples 20:40:00 and on the wiki 20:40:04 I don't think we have octavia in devstack proper yet, right? You still have to manually add the plugin into local* 20:40:18 well, we have a sample config in tree 20:40:27 Yes 20:40:28 our tree 20:40:47 that should be fixed but people are smart enough to not rely on samples :-) 20:41:00 Oh yeah, I forgot about that. 20:41:07 Gotcha, thanks for clearing that up. 20:41:34 cool 20:41:42 they started to remove the table clothes 20:42:02 Hahaha, ok, so time to wrap it up? 20:42:08 yep 20:42:12 bye 20:42:15 bye 20:42:15 bye 20:42:16 octavia has the devstack plugin correct, just enable_plugin in localrc 20:42:20 bai bai 20:42:21 #endmeeting