18:02:57 #startmeeting networking_policy 18:02:58 Meeting started Thu Jun 4 18:02:57 2015 UTC and is due to finish in 60 minutes. The chair is SumitNaiksatam. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:02:59 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:03:01 The meeting name has been set to 'networking_policy' 18:03:16 #info agenda https://wiki.openstack.org/wiki/Meetings/GroupBasedPolicy#June_4th.2C_2015 18:03:36 is ajay or amit here? 18:04:32 yapeng: hi 18:04:45 hello, everyone 18:05:19 sorry for the delay, i was looking for amit, ajay, magesh and ivar 18:05:24 but none of them seem to be on 18:05:41 ah ransari is here, thanks for joining rukhsana 18:05:51 #topic Bugs 18:05:58 you're welcome sumit 18:06:24 we did not have any critical bugs in the last week 18:06:42 however a bunch of new High priority ones got added 18:07:40 #link https://bugs.launchpad.net/group-based-policy/+bugs?search=Search&field.importance=High&field.status=New&field.status=Incomplete&field.status=Confirmed&field.status=Triaged&field.status=In+Progress&field.status=Fix+Committed 18:08:23 more specifically, we ajay (who was planning to join), is one of the first users actually writing to our API, found a bunch of stuff 18:08:33 joined now 18:08:40 sorry confused with UTC time 18:08:44 ah, and there he is, right on cue - Ajay_: hi 18:08:56 amitbose: we see you too, thanks for joining 18:09:02 will get to the packaging in a minute 18:09:17 hello, sorry I'm late 18:09:35 so as i was saying, Ajay_ is one of the first brave new users of the GBP API 18:09:46 ivar-lazzaro: np, we are just getting warmed up ;-) 18:09:53 i am integrating GBP into Rally 18:10:03 Ajay_: yes thanks 18:10:06 tests i mean 18:10:23 so he has been looking at various different aspects as he is working through writing the Rally tests 18:10:41 he has also posted a bunch of bugs, which i think are very good feedback for the team here 18:11:05 so far 5 UI bugs and a 3 were non UI issues 18:11:13 Ajay_: do you feel comfortable pointing out a few of the high priority bugs and/or generall issues that you are seeing? 18:11:35 ya mainly the UI issues with regards to presenting the view to users 18:12:04 Ajay_: but i think some of your observations on the service and API side are more pressing 18:12:18 ya for API side 18:12:25 these issues 18:12:40 https://bugs.launchpad.net/group-based-policy/+bug/1462024 18:12:40 Launchpad bug 1462024 in Group Based Policy "Concurrent create_policy_target_group call fails" [High,Confirmed] - Assigned to Robert Kukura (rkukura) 18:12:56 where concurrent usage of policy target group was causing issues 18:13:30 https://bugs.launchpad.net/group-based-policy/+bug/1460831 18:13:30 Launchpad bug 1460831 in Group Based Policy "API for group update is not clear" [High,Confirmed] - Assigned to Sumit Naiksatam (snaiksat) 18:13:36 for Group update API 18:13:51 and there is one more on treatment of orphaned openstack resources 18:14:01 those sumit where the ones from API 18:14:29 Ajay_: yes 18:14:36 also i was faced with no documentation on python API 18:14:43 right 18:14:53 which makes the development cycle for app users writing applications on top of GBP longer 18:15:04 so as a user, Ajay_ expected that there be a python SDK that he could have used 18:15:07 and we dont have one 18:15:23 Ajay_: that's great feedback! 18:15:34 also i noticed a few cases where i am deviating from intenant 18:15:34 SumitNaiksatam: do other OpenStack projects have a SDK? 18:15:40 this is an area we have to address immediately, so if there’s anyone wanting to work on this, please let me know 18:15:43 ivar-lazzaro: yes 18:16:12 example i still need to specify quoatas in neutron still need to know neutron port to attach A VM etc 18:16:26 #link https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0CC4QFjAA&url=https%3A%2F%2Fwiki.openstack.org%2Fwiki%2FSDK-Development%2FPythonOpenStackSDK&ei=35VwVZcIxfygBPeUgYgC&usg=AFQjCNEl4q5Yu_SQZLQFWQNQlOlp73kIDQ&sig2=KLHnZGp8bQzWCrWb77AN6Q&bvm=bv.95039771,d.cGU 18:16:37 as an App API user i want to perform all operations in intent and dont want to keep track of neutron 18:16:53 Ajay_: i believe we should track that as a bug too 18:17:16 ivar-lazzaro: for the integrated projects, they were part of the SDK 18:17:28 Sumit:Ajay: w.r.t 1460831 - does the bug translate to "Why is a key of provider/consumer required if it is already called out in the CLI via --provided-rule-set?" 18:17:29 ivar-lazzaro: however we can use that SDK as a lib, and build our own 18:17:46 SumitNaiksatam: I found this #link https://github.com/stackforge/python-openstacksdk 18:17:48 ransari: yes 18:17:58 sumit: ok 18:18:00 ivar-lazzaro: yes, thats the one 18:18:43 ransari: sorry, it relates to “what is the use of the scope, and its confusing the way its structured” 18:18:45 Sumit which one quota one or the neutron port knowledge to attach VM what needs a bug? 18:18:58 Ajay_: the quota one 18:18:59 Ajay_: we had a wrapper in the client for VM creation with --group option 18:19:18 Ajay_: for the neutron port, we will address that as a part of the compute integration 18:19:25 also even if we dont have a SDK immediatly documentation on the current python API , Like REST body to use would be useful 18:19:35 not sure if it's part of the current release anymore 18:19:42 Ajay_: agree, we will try to address this one way or the other at the earliest 18:19:49 ivar-lazzaro: its probably not 18:20:04 ivar-lazzaro: i mean its not, and it probably wont be 18:20:26 we cant expect application developers using GBP to capture output from CLI and than put those REST body in their python scripts .... 18:20:41 thats all i had on feedback 18:20:43 for this week 18:20:47 Ajay_: yes, agreed, and apologies for making you do that 18:21:03 currently all APIs are in python language? 18:21:08 Ajay_: great, thanks much (but dont go away) 18:21:13 yapeng: true 18:21:14 yes wont 18:22:01 Ajay: in terms of documentation, is the expectation something along the lines of: http://developer.openstack.org/api-ref-networking-v2.html 18:22:05 yapeng: but we still need an SDK that abstracts the client side usage 18:22:26 ransari: did you just volunteer to do that? :-P 18:22:38 ransari: yes 18:22:38 SumitNaiksatam: makes sense 18:22:54 ajay: ok 18:23:16 sumit: will discuss with hemanth if we can help out with that 18:23:30 the WADL formatting is a bit time consuming to code 18:23:36 hence the delay in doing this 18:23:39 ransari: sure 18:24:25 okay, i know we deviated a little bit from the usual meeting flow today (and perhaps we should called this topic - User Feedback), but i wanted to get Ajay_’s input while its fresh in his mind 18:25:01 #topic Test - Functional, Integration, Rally 18:25:16 note that i added Rally here 18:25:27 so like Ajay_ mentioned he has been working on writing these tests 18:25:45 Ajay_: can you give a quick summary of the scenarios that you are currently targeting? 18:26:10 I am testing all the Implicit Policy APIs for functional, concurrency and scale 18:26:28 to do this i am using the python client API to code these tests in Rally benchmark tool 18:26:51 so far policy actions/calissifiers/rule/ruleset/target group/target are codes 18:26:53 coded 18:27:00 these APIs i mean 18:27:17 Ajay_: why do you say “implicity policy” APIs? 18:27:19 code is in my github under username: akalambu forked rally repo 18:27:30 as in i create a policy target group 18:27:35 and it creates a L2 and L3 policy 18:27:40 Ajay_: ok 18:27:54 you mean the implicity workflows 18:28:04 Ajay_: great! 18:28:11 next would be explicit L2/L3 policy and external network 18:28:21 Ajay_: as regards a home for the tests 18:28:29 Ajay_: these should be in the GBP upstream repo 18:28:46 i can speak to Boris and see if he is ready to accept this upstream? 18:28:51 Boris is Rally PTL 18:29:05 Ajay_: ah, i was thinking this should be part of the GBP repo 18:29:12 other projects do it that way 18:29:16 ok 18:29:26 ya we can do that too 18:30:06 Ajay_: yeah, we can have a “rally” directory here #link https://github.com/stackforge/group-based-policy/tree/master/gbpservice/tests 18:30:20 or in any other location that makes sense 18:30:29 what does the rest of the team think about this? 18:30:35 sumit: issue is it needs rally infra to run 18:30:44 which is in a different rpo 18:30:46 repo 18:30:54 Ajay_: yes agreed 18:31:06 Ajay_: but we have integration tests that need devstack too 18:31:31 seems okay to me, have no issue with it 18:31:36 Ajay_: so we would have a separate job that would setup the rally infra to exercise these tests 18:31:44 igordcard_: ok good 18:31:53 sumit: sounds good will work on i 18:32:15 Ajay_: ok cool, we can connect offline to discuss the feasibility and mechanics of this 18:32:27 sumit: sounds good 18:32:31 any questions for Ajay_ on this? 18:32:35 I would go with as much in-tree as possible at least as far as the tests are concerned 18:32:46 ivar-lazzaro: +1 18:32:59 so that any developer can write basic tests for their new features 18:33:09 who knows... Maybe even as a requirement :p 18:33:27 ivar-lazzaro: integration tests should definitely be a requirement 18:33:34 unfortunately we are not there yet 18:33:53 exercise script is helping in some way 18:34:09 Ajay_: thanks much for your work and update on this as well! 18:34:18 sumit: welcome 18:34:20 i dont have any updates on the functional and integration tests 18:34:26 #topic Packaging 18:35:03 folks, apart from Ajay_ we have a second new member to the team as well, please welcome amitbose 18:35:18 Hello! 18:35:20 amitbose: Welcome! 18:35:30 Hello amitbose! 18:35:31 amitbose is currently focussing on the ubuntu packagind and has made good progress 18:35:36 amtibose: Hi! 18:35:50 rkukura: last week you asked about this, i believe amitbose will be able to give an update 18:35:53 amitbose: over to you 18:35:58 great! 18:36:42 So far I've been looking at build packages from stable/juno branch, but it should be pretty straight forward to port them for kilo 18:36:53 amitbose: ok 18:37:00 amitbose: and we will be posting them on the ppa? 18:37:27 yeah, sure I can do that ... I will need the signing key though 18:37:40 #link https://launchpad.net/~group-based-policy-drivers/+archive/ubuntu/ppa 18:37:48 amitbose: ok lets discuss that offline 18:38:28 amitbose: thanks for the update, looking forward to more in the coming weeks 18:38:42 rkukura: anything at your end for the k-3 packages on fedora? 18:38:56 amitbose: Lets compare notes at some point regareing Red Hat vs. Ubuntu packaging 18:39:13 Hello amitbose 18:39:26 rkukura: yes definitely, had meant to send you an email prior to this meeting on that topic 18:39:35 I have not yet put much effort into updating the Fedora packages to k-3, but expect to get this done in the next couple days 18:39:47 rkukura: ok thanks 18:40:02 ransari: did you have a question for amitbose? 18:40:08 And I’d like to start looking into delorean and how that would impact us 18:40:16 One question: do we need packages only the latest milestone (2015.1.0.0b3) or earlier ones as well? 18:40:19 rkurkura: will the instructions you sent for RDO work on RHEL OSP6 18:40:26 rkukura: okay, but that will not impact us for kilo, right? 18:41:05 probably not immediately for kilo, but I think they are working to support stable branches for k foreward 18:42:11 Ajay_: OSP6 would be similar, except that the GBP packages are not yet in the OSP yum repo, so you’d need to add the RDO repo or some other repo, or install packages explicitly 18:42:54 rkukura: can we capture that detail on the wiki page? 18:43:29 rkukura: can we get the right repo to use...thats always the issue with RHEL 18:43:39 SumitNaiksatam: I’m not sure putting OSP details on the RDO wiki page is appropriate, but we can find somewhere 18:44:00 rkukura: okay 18:44:24 The standard RDO repos work with RHEL, CentOS, and Fedora, but not necessarily with RHEL OSP 18:45:27 rukukura/sumit: lets discuss the RHEL OSP part offline in email 18:45:55 Most likely the RDO RHEL packages will work on the corresponding RHEL OSP release, but there could be dependency issues, and it would be important to make sure RDO packages don’t have priority over supported OSP packages. 18:47:33 Ajay_: rkukura: ok lets take this offline, we are begind schedule 18:47:37 rkukura: thanks for the update on this! 18:47:44 ok 18:48:05 #topic Kilo features 18:48:14 floating ip support: #link https://review.openstack.org/#/c/167174 18:48:34 the exercise script is validating that this is working per the current design 18:48:54 please review and comment if you have objections with this patch going forward 18:48:58 else we need to close on this soon 18:49:26 Service chain driver refactoring: #link https://review.openstack.org/#/q/status:open+project:stackforge/group-based-policy+branch:master+topic:bp/node-centric-chain-plugin,n,z 18:49:46 ivar-lazzaro has been deligently pluggin away on this and updating the patches 18:49:56 and it has shaped up really nicely 18:50:18 so lets provide the review support for this as well 18:50:30 and try to close on this at the earliest 18:50:33 I haven't yet had a chance to calmly review all the latest patches on that but will do it throughout today and this weekend 18:50:41 Hopefully I'll be able to provide a driver for testing it soon 18:50:52 igordcard_: yes being calm is important, its a lot of stuff :-) 18:51:07 igordcard_: thanks! 18:51:21 ivar-lazzaro: my assumption is that we will at least have a heat-based driver to test the basic integration with the neutron services like before? 18:51:36 ivar-lazzaro: so that our exercise scripts can be used to validate that 18:51:58 SumitNaiksatam: I thought magesh is working in something heat based 18:52:08 ivar-lazzaro: okay 18:52:08 SumitNaiksatam: so I thought I would do the Nova driver instead 18:52:34 SumitNaiksatam: which will also make proper use of the plumber and should nicely integrate with igordcard_ work 18:52:37 ransari: songole: so magesh is working on porting the appliance driver for the neutron services? 18:52:48 ivar-lazzaro: okay 18:52:59 yep 18:53:03 what is the current date for the final kilo-gbp? https://etherpad.openstack.org/p/kilo-gbp-plan seems to be outdated 18:53:08 SumitNaiksatam: not sure. will find out. 18:53:17 igordcard_: you need to rebase your patch? 18:53:34 igordcard_: we are looking at Kilo release for the end of this month latest 18:53:45 SumitNaiksatam: started reviewing the patches. Hope to complete over the weekend. 18:53:53 SumitNaiksatam: probably, I had already rebased it once, but it should be no problem in its current state 18:53:58 so we will try to get into the RC phase in about 10 days time frame 18:54:09 songole: okay 18:54:19 SumitNaiksatam: okay, thanks, will write that down on calendar 18:54:44 which means that we need all this reviewed at the earliest 18:55:26 any thoughts/comments/objections on either of the above features being discussed? 18:56:04 Magesh will start work on porting appliance to service-node model sometime next week 18:56:05 ivar-lazzaro: thanks for the update and the fantastic work so far! 18:56:15 ransari: next week??? 18:56:25 ransari: lets discuss offline 18:56:47 There was one more feature I wanted to discuss, but I guess I'll send out an email instead 18:56:53 we need to be able to validate that all this works end-to-end, and without a driver how do we do that? 18:57:04 sumit: once service refactor review is cmpletd and merge to kilo is done 18:57:06 ivar-lazzaro: ok, you can summarize here 18:57:40 sumit: we can discuss offline 18:57:53 Given how many operations we issue in the backend (postcommit), I believe it's critical for usability to move our project towards a promise theory approach 18:57:58 ransari: i would have preferred to see the driver in place in parallel, since the service refactor work does include the refactored reference driver 18:58:12 ivar-lazzaro: okay 18:58:14 But even before doing that, having an asynchronous backend is critical 18:58:25 ivar-lazzaro: yes, okay so thats a longer discussion ;-) 18:58:38 SumitNaiksatam: it is :) 18:58:45 ivar-lazzaro: Do you feel TaskFlow is up to the job for this? 18:58:56 ivar-lazzaro: thanks for bringing it up 18:59:01 rkukura: definetively, but I think we need a solution for Kilo 18:59:06 rkukura: i would assume we use it to realize the async 18:59:29 ivar-lazzaro: hmmm, kilo might be challenging, but perhaps you have thought through this 18:59:49 lets discuss offline until next meeting 18:59:49 rkukura: otherwise usability will not be great (or even fair) 18:59:59 ok 19:00:02 #topic Liberty features 19:00:28 i trust the other folks - yapeng, yi are working in parallel on the kilo features 19:00:51 we will get spending more time on those once the kilo backlog is cleared 19:01:12 okay so we are out of time for today 19:01:16 any parting thoughts? 19:01:37 ivar-lazzaro: Lets discuss async/TaskFlow tomorrow if possible. 19:01:39 liberty features: probably that drag-and-drop SC UI, but for kilo most likely not 19:01:45 rkukura: yes! 19:02:07 SumitNaiksatam: yes, the renaming! 19:02:15 #link https://etherpad.openstack.org/p/gbp-rename-proposals 19:02:18 igordcard_: yes that for kilo 19:02:28 ivar-lazzaro: yes, thanks for the reminder 19:02:36 here are a bunch of proposals 19:02:45 I'm very sad about "Meta" and 19:02:49 "Regi'a 19:02:54 being unusable 19:03:11 I still like intento 19:03:11 I'll start a poll today 19:03:15 ivar-lazzaro: cool 19:03:31 ivar-lazzaro: lets hold on the poll 19:03:35 hopefully we'll have a decision by next week (or before Kilo closes) 19:03:42 SumitNaiksatam: why is that? 19:04:19 ivar-lazzaro: give people some time to thing 19:04:21 think 19:04:29 Why not rename in liberty at the time we become a standalone server? 19:04:33 (by the way I wonder who proposed "Membrane"... It reminds me of something :) ) 19:04:34 i would prefer to get people focussed on getting kilo done first 19:04:45 SumitNaiksatam: fair enough 19:04:50 ivar-lazzaro: i proposed membrain 19:04:58 note, its not membrane ;-) 19:05:11 SumitNaiksatam: lol :D! was the Embrane reference intended? 19:05:12 rkukura: yeah 19:05:25 ivar-lazzaro: it might seem like, but no :-P 19:05:41 alright, lets wrap it up for today 19:05:42 SumitNaiksatam: the company name actually derives from membrane... But we digress 19:05:53 ivar-lazzaro: i would image :-) 19:05:55 thanks all! 19:05:58 bye 19:06:00 bye 19:06:03 adieuuu 19:06:05 bye all 19:06:08 #endmeeting