00:01:13 #startmeeting congressteammeeting 00:01:14 Meeting started Thu Aug 24 00:01:13 2017 UTC and is due to finish in 60 minutes. The chair is ekcs. Information about MeetBot at http://wiki.debian.org/MeetBot. 00:01:15 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 00:01:17 The meeting name has been set to 'congressteammeeting' 00:01:50 hello all! 00:01:56 hello 00:01:57 topics here as usual: https://etherpad.openstack.org/p/congress-meeting-topics 00:02:11 please read/add/comment =) 00:02:44 hello masahito 00:02:48 Hi all 00:02:54 hello ramineni_ 00:03:14 masahito: how was the event in mexico? 00:03:43 we had a good discussion. 00:04:38 Hi all 00:04:45 hi thinrichs 00:04:57 masahito: great! 00:05:27 ok let’s dive into it then. not too much to talk about today. 00:05:35 #topic pike 00:05:53 final release is coming out next week. this is our last chance to release a new RC if needed. 00:06:59 This is the only new bug I’m aware of since RC1. https://bugs.launchpad.net/congress/+bug/1712177 00:07:00 Launchpad bug 1712177 in congress "multi-proc devstack sometimes fails on encryption key" [Medium,In progress] - Assigned to Eric K (ekcs) 00:07:13 And it doesn’t seem to warrant a new release. 00:07:53 any thoughts or comments on pike? 00:09:01 That bug only effects testing? 00:09:12 Or does it impact a real multi-node deployment? 00:10:06 Primarily testing, but it could come up in real deployment in very specific circumstances. 00:11:02 deploying multiple congress instances on the same server doing first launch at the same time. 00:11:09 I guess the real fix would be retrying a while depending on the failure 00:13:43 but even in failure (can occur only on very first launch, not possible subsequent restarts managed by agent), one just needs to retry. 00:14:13 That seems like it's not a big deal then. The retry can happen outside Congress (for now). 00:14:53 the fix I submitted simply skips the encryption key initialization step on any node that doesn’t host datasource. Since we only ever support one datasource instance at a time that fixes the issue. 00:15:05 #link https://review.openstack.org/#/c/495989/ 00:15:25 If in the future other types of services also need access to the key then we can do something else. 00:15:29 thinrichs: agreed. 00:16:13 anything else on pike? 00:17:19 Not from me. 00:18:17 ok then moving on. 00:18:26 #topic PTG 00:19:10 PTG is coming up 9/11-9/14 00:19:23 topic brainstorm document is here: #link https://etherpad.openstack.org/p/congress-ptg-queens 00:20:20 please be thinking and jotting down thoughts so we can have a great productive discussion at denver =) 00:21:34 concrete use cases and requirements are particularly helpful for driving development direction. 00:22:30 I’m working to gather them from people. more is always better =) 00:22:58 any thoughts or comments on topic brainstorm? 00:23:13 Sounds good. I think use cases and customer pain-points are key. 00:25:06 yups. ok another PTG thing is just logistics. some have asked about remote option and I’m going to try for that. 00:26:09 there is a section in the planning etherpad for conferencing solution suggestions. If you have experience with something that worked well please suggest it here or on the etherpad. 00:27:21 I’m hoping to try out at least one before the PTG. 00:28:12 ok any thoughts on the PTG? 00:28:13 Hangouts, bluejeans, zoom, webex are the ones I'm most familiar with. 00:28:24 Uber has one too that I was impressed by. 00:28:35 The only one I know is free is Hangouts 00:28:56 gotomeeting is one of the options. 00:29:42 great. I think I have access to host on webex though the company. 00:30:09 bluejeans has a one month trial. but even if not it’s not expensive and only the host needs to pay. 00:30:32 any comments on which ones have worked better than others in your experience? 00:30:53 gotomeeting allows three user to join it in free plan. 00:31:31 Forgot about that one 00:33:31 ok great. I guess I’ll do some digging and find an acceptable one. 00:34:11 I might need to bring two laptops to have one capturing the whiteboard haha. 00:35:36 ok let’s move on then. 00:35:40 #topic open discussion 00:36:10 anything else we wanna talk about? not super often all four of us are here hehe. 00:37:36 I don't have anything pressing 00:37:55 nothing from my side. 00:38:55 One thing I’ve been thinking about for quite a while now is whether we want Congress to deal with mission-critical, latency-critical situations. 00:40:09 An example is something like the recovery scenarios masahito_ has worked on. 00:41:25 At the moment, it’s hard to Congress to promise very low latency in rule firing partly because of just engineering and partly because of the nature of rule engines. 00:42:24 so a question that comes up is whether there are compelling and suitable enough use cases is that space to try and engineer that into congress. 00:43:03 one question is how uniform the desired behavior is from deployment to deployment. 00:44:17 if the desired behavior is fairly uniform and rarely changed, then it seems the right solution is dedicated code to handle the behavior. Maybe with just a few configuration options to control the behavior. 00:44:45 I'd want to know the use-cases and how low-latency they are. 00:44:58 And whether the eventual-consistency model we have for gathering data is adequate 00:45:28 but if there is enough variety needed from deployment to deployment, especially if human interaction with the policy is frequent, then a Congress style model could make more sense. 00:46:10 So I just wanted to throw it out there to get one people’s minds. 00:46:14 thinrichs: yes absolutely. 00:46:43 masahito: do you have any comments on latency requirement and model for use cases you’ve been involved in? 00:47:00 yes. 00:47:15 if I remember correctly from the presentations we’re looknig for 1s at the worst end to end, but preferrably much better? 00:48:46 Congress passed the requirements in our current tests with small scale deployment. 00:49:21 like 200 instances on 10 hypervisors. 00:50:48 Got to run. See you all later! 00:51:04 masahito: got it. so in those requirements, roughly how much time does congress have to react? 00:51:12 bye thinrichs ! 00:51:51 haha masahito_ connection problems? 00:53:56 haha maybe we should continue another time. 00:55:33 sorry, the network is not unstable today.... 00:55:49 s/is not/is/ 00:55:52 haha np. maybe we should continue another time. 00:56:00 time almost up anyway. 00:56:12 yup. 00:56:45 alright let’s end meeting then unless someone has something to say. 00:57:36 ok bye everyone! 00:57:46 #endmeeting