00:01:56 <thinrichs> #startmeeting CongressTeamMeeting 00:01:57 <openstack> Meeting started Thu Jun 30 00:01:56 2016 UTC and is due to finish in 60 minutes. The chair is thinrichs. Information about MeetBot at http://wiki.debian.org/MeetBot. 00:01:59 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 00:02:01 <openstack> The meeting name has been set to 'congressteammeeting' 00:02:09 <aimeeu> Hi! 00:02:16 <ekcs> hi all 00:02:16 <thinrichs> Hi all 00:02:35 <masahito> hi all! 00:03:21 <thinrichs> ramineni can't attend today 00:03:28 <thinrichs> So I guess we're all here 00:03:37 <thinrichs> Agenda is... 00:03:44 <thinrichs> 1. Barcelona items 00:03:47 <thinrichs> 2. Status 00:03:52 <thinrichs> Anything else? 00:04:16 <ekcs> nothing from me. 00:04:25 <masahito> me too 00:04:38 <thinrichs> #topic Barcelona 00:05:35 <thinrichs> Did folks receive their discount code for registration? 00:05:43 <aimeeu> Yep 00:05:59 <masahito> yes 00:06:12 <ekcs> yup. I’m signed up. 00:06:59 <thinrichs> Great. Just wanted to make sure everyone had what they needed. 00:07:09 <thinrichs> Also, talk submissions are now open 00:07:21 <thinrichs> Does anyone have ideas for submission? 00:09:26 <thinrichs> I feel we've been so focused on HA, etc. that we haven't spent much time on new applications that people would want to hear about 00:09:46 <thinrichs> masahito: your talk in Austin was good for that reason. People could see what Congress is useful for. 00:10:06 <masahito> oh, thanks 00:10:28 <thinrichs> Well, let's all think about it 00:11:11 <masahito> yap. 00:11:29 <thinrichs> #topic Status 00:11:37 <thinrichs> masahito: want to start? 00:11:46 <masahito> yes. 00:12:27 <masahito> I have 2 updates. 00:13:05 <masahito> 1. last 2 week I was in OPNFV summit and I prepared the demo in the summit. 00:14:02 <masahito> the slides about the talk and demo are available in here 00:14:11 <masahito> #link https://wiki.opnfv.org/display/doctor 00:15:00 <masahito> oh, OPNFV is one of usecase for congress I'm working for now. 00:15:25 <thinrichs> Cool! How'd the demo go? 00:15:42 <ekcs> nice! btw masahito I talked to wenjing about Conrgess and NFV because of your contact. 00:15:47 <masahito> It was nice. 00:15:58 <aimeeu> masahito: are you familiar with OPNFV's Copper project? 00:16:06 <masahito> ekcs: yeah, I met him. 00:16:54 <masahito> aimeeu: I know the project name but I'm not good at the project now. 00:17:30 <aimeeu> masahito: I thought Copper integrated Congress into OPNFV 00:17:45 <masahito> 2. I updated the patch for race condition of datasource creation 00:18:33 <masahito> that's from my side 00:19:56 <thinrichs> masahito: thanks! 00:20:49 <thinrichs> aimeeu: last I heard Copper was doing gap-analysis and putting people on existing projects to fill those gaps 00:20:56 <thinrichs> but it could have changed over the last year for sure 00:21:13 <thinrichs> ekcs: want to go next? 00:21:20 <aimeeu> thinrichs: bryan_att did a presentation on Copper with Congress at the OPNFV Summit last week. 00:21:32 <masahito> aimeeu: yeah. I saw a talk that said Copper has already implemented Congress installer and had some concreat policy rules. 00:22:43 <thinrichs> aimeeu: what did the audience think of the talk? 00:23:25 <aimeeu> Unfortunately I wasn't there and haven't had a chance to actually talk to bryan_att, who is out of town on business again this week (and next week he's on vacation) 00:24:13 <aimeeu> bryan_att's slide deck is on the Copper page: https://wiki.opnfv.org/display/copper 00:25:53 <thinrichs> Cool! I'll have to take a look. 00:26:02 <thinrichs> Anything else on the OPNFV summit to discuss? 00:26:05 <ekcs> great I guess we can talk to bryan_att when he’s around. 00:27:09 <thinrichs> ekcs: want to do a status update? 00:27:17 <ekcs> thinrichs: yup. 00:27:23 <ekcs> 1. Completed HA-spec up. #link https://review.openstack.org/#/c/318383/ 00:27:23 <patchbot> ekcs: patch 318383 - congress-specs - Support HT-HA deployments 00:27:31 <ekcs> 2. Been manually looking through code to annotate things having to do with thread-safety. WIP here: #link https://review.openstack.org/#/c/335721/ 00:27:31 <patchbot> ekcs: patch 335721 - congress - WIP - Thread-safety analysis 00:27:44 <ekcs> It's a bit complex because so many things are blocking now. It's almost like the standard greenthread assumption that execution is not interrupted unless there is a block/yield is not very useful any more. I'll continue on and see. 00:27:48 <ekcs> One question I haven't been able to find the answer to: does importutils.import_module(...) block? It seems like some imports block and others don't. Pls lmk if you happen to know the answer. 00:28:00 <ekcs> That’s all from me. 00:31:32 <thinrichs> Those annotations look scary. So many are blocking 00:31:39 <thinrichs> I guess in the API that's expected 00:31:46 <thinrichs> since it's basically just a message router 00:32:48 <thinrichs> Looks like the policy engine is pretty clean. Thankfully 00:33:36 <ekcs> thinrichs: yea that’s expected in API. but in general lots of blocking everywhere. after some more work i’ll have a better picture of whether we should expect to be able to stamp of thread safety things with the current approach. 00:34:06 <ekcs> thinrichs: hopefully, tho I haven’t finished agnostic =| 00:35:11 <thinrichs> I would expect agnostic and the datasources to be pretty clean, because of our pub/sub model. 00:35:32 <ekcs> thinrichs: agreed. 00:35:37 <thinrichs> Those bits aren't actively sending messages (at least in the code WE write) 00:35:44 <masahito> I think some of methods in agnostic would be replaced with pub/sub call in the HA model. 00:36:01 <thinrichs> masahito: I didn't think we were changing much in the HA model 00:36:08 <thinrichs> for agnostic 00:36:20 <thinrichs> (That's a question I asked in the work items on the spec, BTW) 00:36:39 <thinrichs> Is agnostic needing to change for HA? 00:37:45 <masahito> I thought local leader election needs a bit replacing. 00:38:12 <thinrichs> I thought local leader was taken care of by the datasources 00:38:35 <thinrichs> ekcs: ^ 00:39:11 <ekcs> right local leader logic is isolated in DSD. 00:39:38 <thinrichs> ekcs: does agnostic need to change (at all) for HA? 00:39:40 <masahito> sorry, I was confused at we had some DSD for same datasource. 00:40:04 <ekcs> I expect most changes won’t need to touch agnostic. But rather made in the base classes. Not totally sure I can spend some time thinking through it. 00:41:04 <ekcs> it really comes down to which bits remain in agnostic.py. 00:41:39 <ekcs> main changes: sending execute_action (should be done in wrapper). 00:42:21 <ekcs> understanding joining a runnnig deployment. (should be done in wrapper?) 00:43:26 <thinrichs> Does execute_action need to change at all? I would have thought it just sends the action to the name of the datasource as always. The datasource needs to change b/c it now gets 3 of those (1 from each PE), and ignores 2 of them. 00:43:34 <ekcs> remembering action execution requests for not missing action. that’s an agnostic change. 00:44:21 <ekcs> thinrichs: I think the only change really needed is to make sure the receiving DSD can tell who the sender is. Not totally sure if that’s already in the infra or not. 00:44:46 <thinrichs> That sounds right to me, at least for version 0.1 of HA. Other stuff might need to be added later on. 00:45:13 <ekcs> I can spend some time teasing out which PE changes are in the dse level and which PE changes are an the agnostic level. 00:45:27 <ekcs> it depends on the details a little bit. 00:46:19 <thinrichs> I put a list of what I think we'd need in one of my comments on the spec. It didn't seem like there was a ton to do, actually. 00:46:37 <thinrichs> At least if we want 'at-most-once' semantics for action execution 00:46:57 <thinrichs> i.e. just leader election on the DSDs to eliminate dups 00:47:16 <thinrichs> So maybe take a look at that and see if it sounds right. 00:47:24 <thinrichs> 13 min left 00:47:28 <ekcs> yup the substantive things are not that many. 00:47:50 <thinrichs> aimeeu: want to give us a status update? I see you pushed some code 00:48:30 <aimeeu> Sure. I've been trying to learn the code and learn Python at the same time, so my progress has been pretty slow. 00:49:11 <aimeeu> I added the execute(...) method to PlexxiDriver and pushed the code. Thanks for the feedback. Pulled my hair out trying to get a unit test to work. 00:49:53 <thinrichs> That one was especially nasty. I couldn't tell what was wrong until I looked at the errors 00:50:14 <thinrichs> It was definitely tough for someone who hadn't worked on the datasources before 00:50:14 <aimeeu> I'm getting more grey hair by the day! Today I was playing around with the CLI and command line to methods in the python client. Also picked up a bug there. 00:50:43 <thinrichs> Cool! 00:50:52 <aimeeu> It's challenging but fun. My real problem now is I haven't been able to get devstack to stack. Bah! 00:51:15 <ekcs> thanks aimeeu! 00:51:17 <thinrichs> That's a common problem. 00:51:39 <aimeeu> bryan_att set me up on one of his servers so I can play around with a Congress installation. 00:51:41 <masahito> aimeeu: Great, I found some tiny bugs. do you need these? 00:51:50 <ekcs> aimeeu: yea something always goes wrong. Feel free to ping me on irc with osmething it might be something I’ve seen. 00:51:58 <aimeeu> masahito: Yes please!! 00:52:16 <masahito> #link https://bugs.launchpad.net/congress/+bug/1597243 00:52:16 <openstack> Launchpad bug 1597243 in congress "Policy Detail API returns 500 status codes in dse2" [Undecided,New] 00:52:16 <aimeeu> ekcs: Great - I'll keep you in mind and try not to be a pest. ;) 00:52:25 <masahito> #link https://bugs.launchpad.net/congress/+bug/1597555 00:52:26 <openstack> Launchpad bug 1597555 in congress "keystone_driver fails to poll user data" [Undecided,New] 00:53:15 <aimeeu> masahito: Excellent - thanks for those! 00:53:16 <thinrichs> aimeeu: if you're using irc, definitely put our handles in so we get notified. There's not a ton of traffic on there usually, so if I don't get a notification I'll probably miss your request. 00:53:25 <masahito> I think the 2 is not so difficult bugs :) 00:54:02 <masahito> s/the 2 is/ the 2 bugs are/ 00:54:23 <ekcs> aimeeu: on the CLI maxwidth issue, it might be something on openstack CLI end not congress-client end. Just saying be aware that it may not be solvable. 00:54:34 <aimeeu> thinrichs: I noticed the IRC congress channel is pretty quiet, so I was wondering what the best way of contacting people if I have questions - I'll definitely put your handles in 00:55:05 <aimeeu> ecks: thanks for the tip. or it might be in the 3rd party library that formats the output 00:56:07 <aimeeu> cliff – Command Line Interface Formulation Framework is the library that formats output 00:56:10 <thinrichs> FYI… The 503 bug is probably fixable just within Congress. Finding a problem in the keystone datasource will require finding a mismatch between the way the keystone client works and the way the Congress keystone datasource is using it. 00:56:47 <thinrichs> aimeeu: I'd try the IRC, and if that doesn't work, use the ML. 00:57:01 <aimeeu> thinrichs: thanks - I'll keep that in mind when I start work on that bug. 00:57:21 <thinrichs> We're spread out across the globe..masahito is in Japan, ramineni is in India, and ekcs and I are on the west coast in the US 00:57:26 <ekcs> I’ll do a better job of being on IRC. It’s just easy to forget because my workplace blocks IRC =| I’m also reachable over email. 00:57:52 <thinrichs> 3 minutes left 00:57:57 <thinrichs> Quick status update from me. 00:58:27 <thinrichs> I pushed some patches on top of masahito's to get Congress running in 2 separate processes 00:58:32 <thinrichs> Seems to be working fine 00:58:46 <thinrichs> 1 server is PE + API, and the other server is DSD 00:59:03 <thinrichs> Manual tests from the CLI show that everything works as you'd expect 00:59:28 <thinrichs> To try it out yourself, there are instructions in the commit msg 00:59:29 <thinrichs> https://review.openstack.org/#/c/307693/ 00:59:29 <patchbot> thinrichs: patch 307693 - congress - Enable different deployment configurations for HA 01:00:21 <thinrichs> Out of time. I'll hangout in #congress for a few minutes if anyone wants to continue. 01:00:33 <aimeeu> Thanks everyone 01:00:43 <thinrichs> Thanks all! 01:01:09 <masahito> thanks 01:01:23 <thinrichs> #endmeeting