00:00:17 #startmeeting CongressTeamMeeting 00:00:18 Meeting started Thu Nov 19 00:00:17 2015 UTC and is due to finish in 60 minutes. The chair is thinrichs. Information about MeetBot at http://wiki.debian.org/MeetBot. 00:00:19 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 00:00:21 The meeting name has been set to 'congressteammeeting' 00:00:48 Are we all here or should we wait a couple of minutes? Who is here? 00:00:53 hi 00:00:54 hi 00:01:13 hi 00:01:41 hi 00:01:58 Hi all. We have 6 or so here, so let's get started 00:02:33 o/ 00:03:05 My agenda: devstack, distributed arch, python3, and a couple others 00:03:14 hi 00:03:19 I see fabiog and joonmyung are new here, I believe. 00:03:30 yes. hi. 00:03:30 We typically have newcomers introduce themselves at the end of the meeting 00:03:35 thinrichs: yes 00:03:44 But of course feel free to jump in throughout. 00:03:52 thinrichs: no problem. 00:03:54 Anything else for the agenda? 00:04:13 I have a mid-cycle meeting question 00:04:27 is there any plan? 00:04:48 we can talk at the end 00:05:04 That's an open-ended one, so I'll add that to the agenda in the middle 00:05:24 #topic devstack 00:05:38 Remember that the way we currently use devstack in the gate will no longer work at Mitaka1 00:05:43 So we're moving to the devstack plugin 00:05:58 There's been a bunch of back and forth on the patch that ramineni put together 00:06:07 ramineni: want to update everyone? 00:06:13 #link https://review.openstack.org/#/c/241969/ 00:06:31 thinrichs: yes, its ready from our end , but havent been approved yet by infra :( 00:07:05 I've pinged infra twice, but they may have thought we were still discussing 00:07:13 Now that all the reviews are +1 I'll ping them again 00:07:39 thinrichs: ok, ill also ping them for review 00:07:55 thinrichs: im also checking on disabled tempest tests 00:08:09 Once we get that patch in, there's another gate job (from murano) that we may need to touch 00:08:15 Or maybe the murano team will take care of it 00:08:32 ramineni: thanks for taking care of this patch 00:08:41 ramineni: any progress to report on the tempest tests? 00:09:15 thinrichs: yes, some tests are failing because of some recent changes in tempest , so could get them worked 00:09:31 thinrichs: i have still bunch of them to resolve 00:09:40 :) , will be working on it 00:09:56 ramineni: sounds good 00:10:16 #topic distributed architecture 00:10:31 pballand: how's that DseNode patch coming? I saw you pushed to review today 00:11:02 I’ve thrown together a very lightweight skeleton that shows the problem I’ve been stuck on 00:11:10 https://review.openstack.org/#/c/247232/ 00:11:20 RPC works fine on AMQP, but doesn’t on Fake 00:11:33 if anyone has cycles to take a look, help would be much appreciated 00:12:17 problem can be reproduced by running ‘python test_dsenode.py —ampq’ (works) and ‘python test_dsenode.py --fake’ (fails) 00:13:05 I have a bunch of other code prototyped and ready to merge in, but it all relies on RPC actually working :-/ 00:13:31 that’s my update 00:13:38 Which test fails? What's the message? 00:14:09 Or maybe we can do that offline. 00:14:15 test_dsenode.py line 128 00:14:19 Sorry—got too caught up wanting to hunt down the bug 00:14:29 RPC calls don’t return when using the fake driver 00:15:36 sure - in dsenode.py, an rpc server is started on line ~64, and calls are invoked on line ~108 - there isn’t much to the code 00:17:30 Also on the distributed arch front it looks like we just have 1 outstanding API model to get merged 00:17:39 That's gone through a few rounds of review, so should happen soon. 00:18:38 I think that once we get dsenode in, there'll be a bunch of work to do hooking everything up. 00:19:04 thinrichs: still datasources one also left 00:19:14 ? 00:19:16 I'll try to spend some time this week thinking through what needs to be done. 00:19:52 ramineni: ok I'll take a closer look at that 00:20:07 ok 00:20:25 #topic python3 00:20:34 ekcs: how is python3 conversion going? 00:20:47 Making progress on getting all the datalog bits Python3 compliant. Most of the issues have to do with Python3 being more resrictive on what types are used. 00:20:48 It's going rather slow because a lot of time is spent digging through the code to understand the intended behavior in Python2. On the plus side, it's helping me understand the innards of the datalog code, and occasionally exposing a bug. 00:21:20 Any estimates on how much longer? Number of errors, perhaps? 00:22:48 about 30 errors involving datalog. but some tests are skipped due to other errors so who knows. 00:23:19 ekcs: thanks for the update 00:23:21 I’m hoping to have the datalog bits done sometime next week. 00:23:32 ekcs: need any help from us? 00:23:42 but it’s hard to say. some things take 10 minutes. some things take 2 days. 00:24:05 maybe 1 day. 00:24:42 thinrichs: mostly just responding to questions about exsting code, as you guys have been doing. 00:24:42 ekcs: Let us know if you need something 00:24:49 sounds good 00:25:46 #topic mid-cycle meetup 00:26:12 fabiog: did you have something to propose for this? 00:26:23 thinrichs: yes, Cisco is happy to host it 00:26:45 thinrichs: I was thinking to do it in San Jose, CA unless there is a wish to do it elsewhere 00:27:10 thinrichs: I am also trying to coordinate with Monasca mid-cycle so we could have 1 day of joint session 00:27:15 thinrichs: how does it sound? 00:27:22 fabiog: thanks for the offer to host! 00:27:35 thinrichs: my pleasure 00:28:13 I know a number of people are local to San Jose, so that's convenient 00:28:25 But a number of people aren't local to San Jose, so it's not so convenient for them 00:28:52 Maybe we should discuss what topics we're interested in, so we get a feel for who would be most interested in attending 00:29:07 And then discuss the location 00:29:19 thinrichs: also what are historically the datas you have held such a meetings in the winter cycles? 00:30:01 fabiog: not sure about dates 00:30:31 thinrichs: have you guys had a mid-cycle before? 00:30:43 fabiog: a couple 00:30:58 Last cycle we discussed the new architecture 00:31:09 One cycle we had a general policy meet-up 00:31:19 So maybe just those. 00:31:26 pballand: were there any others? 00:31:51 that’s all I recall 00:32:06 I'd propose one topic for this cycle's meet-up is a follow-on to the distributed arch 00:32:28 Either it could be a code sprint to finish things up, or we could discuss the next step: using the distributed arch to do HA and high query throughput 00:32:44 Are there other proposals? 00:33:21 I would like to share a design/architecture using Monasca Alarming and WebHooks to get alerts on policy infringments 00:33:37 or potential infringments 00:34:01 fabiog: that sounds cool 00:35:14 A generalization of fabiog's might be… 00:35:23 deeper integration with other OpenStack projects. 00:35:56 In addition to Monasca integration, we could discuss su_zhang's recent experiments with gating Nova on Congress and whether generalizing that to other services makes sense. 00:36:27 Any opinions? 00:36:58 thinrichs: If there is any space, I'd like to share policy graph abstraction using congress with some current limitations and possible improvements. 00:38:25 thinrichs: if time allows, the more general topic makes sense. but if we don’t have enough time then it may dilute it to the point of not actually making any progress. 00:38:26 joonmyung: is policy graph abstraction a new UI you're proposing for Congress? 00:39:03 ekcs: agreed 00:39:09 thinrichs: Patially, it will be a new UI but it is more than UI. 00:40:12 thinrichs: I am working with JK for this topic. we are using congress for communicating with infra now. 00:40:56 thinrichs: I am looking for some contribution points for your congress now. thanks. 00:41:24 joonmyung: maybe it'd be good to have you do a demo/presentation earlier than the meet-up 00:42:01 joonmyung: the mid-cycle meetup isn't a conference where people give talks; it's typically more of a chance for us to all get in a room and hammer out some design 00:42:45 thinrichs: I see. 00:43:46 Let's all think about topics for the mid-cycle meet-up and discuss again next week. 00:44:59 fabiog: Thanks again for offering to host. I think it'd be great to have Cisco host this time, assuming San Jose works. 00:45:16 thinrichs: yes, no problem 00:45:48 thinrichs: regarding dates maybe mid to late Jan, would that work? 00:46:07 fabiog: sounds like the right timeframe to me. What about everyone else? 00:46:15 I can create a doodle and open it for votes next week 00:47:01 thinrichs: last variable, how many days? 2 or 3 or 4? 00:47:17 fabiog: a doodle poll would be great! 00:48:40 fabiog: not sure how many days; would depend on the topic(s). Maybe poll for 4 00:48:56 That does it for my agenda. Let's switch over to open discussion. 00:49:02 #topic open discussion 00:49:23 fabiog and joonmyung: want to introduce yourselves and tell us why you're interested in Congress? 00:49:28 <_RuiChen> I have a question about API 00:49:44 <_RuiChen> policy/rule status API in status_model.py and execute_action in datasource_model.py exsit in code, 00:49:52 <_RuiChen> but these don't present in API doc, do we actually support these API? 00:50:17 thinrichs: please go ahead and respond to _RuiChen 00:50:29 <_RuiChen> fabiog: sorry 00:50:32 I can introduce myself afterwards ;-) 00:50:39 thinrichs: I agree. 00:50:44 _RuiChen: today there's not much use for /policy/rule/status 00:51:14 _RuiChen: I think we want execution action in datasource though, as a way of testing out actions. 00:51:19 <_RuiChen> thinrichs: yes, all the response present in policy/rule show, 00:51:38 _RuiChen: so I'd remove the policy/rule/status code and update the API doc for execution action on the datasource 00:52:00 <_RuiChen> thinrichs: got it, I can help to do it 00:52:11 _RuiChen: that'd be great! 00:52:31 fabiog: how about that intro? 00:52:35 <_RuiChen> fabiog: joonmyung: please go on :-) 00:52:47 sure, I am Fabio Giannetti and I work for Cisco 00:53:22 I am Monasca Core and I am interested in integrating Congress with Monasca to enable monitoring feedback for real-time push-model policy evaluations 00:53:32 done :-) 00:53:36 :) 00:53:49 fabiog: nice to have you join the meeting! 00:54:04 Ok. I am Joon-Myung Kang and currently working for Hewlett Packard Labs. I've been mostly working in Neutron for policy management, service chaining, and performance guarantee. 00:54:40 Why I am interesting in Congress is that I would like to apply network policy management to more general policy management such as Congress. 00:55:02 <_RuiChen> "feedback for real-time push-model policy evaluations", it's great, welcome 00:55:11 joonmyung: glad you could be here! 00:55:35 thinrichs: doodle link for mid-cycle: http://doodle.com/poll/iinvncz7298rq9wb 00:55:55 fabiog: super fast! Thanks! 00:56:01 thinrichs: thank you. actually, I was in the Vancouver design summit with JK for presenting PGA work. I am mostly implementing it on OpenStack. 00:56:15 Anyone else have questions or anything to discuss in the last 4 minutes? 00:57:08 thinrichs: I still didn't manage to get the datasource driver to store data in the database 00:57:25 I have something weird maybe i’ll bring up in case someone happens to know something 00:57:30 if I create the exact same structure by hand it works, but when I get the data from Monasca it does not store it 00:57:31 You pulled out everything and started from scratch? 00:57:38 thinrichs: yes 00:57:42 I pushed the spec for Push type datasource driver. please review it. 00:57:52 ekcs: go ahead and interleave your question here, or we won't get to it 00:57:53 https://review.openstack.org/#/c/246742/ 00:58:09 thinrichs: I will give it another shot, but I will ping you again if I fails since I am getting out of ideas 00:58:12 fabiog: that's bizarre 00:58:27 masahito: Thanks! I'll take a look. 00:58:29 thinrichs: I konw, you remembered when we did Webex that was really strange 00:59:05 fabiog: what if you print the data that monasca is returning to the screen, copy and paste it into the code. Does that work? 00:59:14 In a test I added, it passes if I run tox -epy34 first and then tox -epy27. 00:59:27 thinrichs: yes, that seems to work 00:59:27 but it often fails if I go straight to epy27. 00:59:42 I didn't copy and paste but I reproduced the structure and worked 00:59:49 I will try to copy and paste 01:00:00 ekcs: that's crazy. 01:00:11 line 737 of this test. https://review.openstack.org/#/c/246042/3/congress/tests/datalog/test_compiler.py 01:00:15 ekcs: those should have separate virtualenvs and shouldn't collide 01:00:22 fabiog: sounds good. 01:00:33 We're out of time, so we can continue in #congress if anyone wants to 01:00:36 #endmeeting