16:02:59 #startmeeting tacker 16:03:00 Meeting started Thu Sep 17 16:02:59 2015 UTC and is due to finish in 60 minutes. The chair is sridhar_ram. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:03:01 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:03:03 The meeting name has been set to 'tacker' 16:03:11 trozet: hi 16:03:19 #chair bobh 16:03:20 Current chairs: bobh sridhar_ram 16:03:47 #topic Announcements 16:04:07 Today's agenda is here - https://wiki.openstack.org/wiki/Meetings/Tacker#Meeting_Sep_17.2C_2015 16:04:23 anything else to discuss beyond that ? 16:05:16 Hi Everyone 16:05:26 Rajkumar_: hi! 16:05:37 few general announcements ... 16:06:00 our first functional test landed in Tacker - thanks a lot santosh! 16:06:33 now you can run 'tox -e functional' on a running tacker devstack to exercise these tests 16:06:53 more tests are expected to be added over time...! 16:07:01 nice 16:07:02 yey! 16:07:47 I'm working w/ infra team to add a gate-job to hook them up to the gate... 16:08:23 initially it will be non-voting .. meaning it won't block a merge but the core team would expect the tests to pass 16:08:34 sridhar_ram: do we need to write a post_hookup script for the job or is just configuration in project-yaml? 16:08:45 once we gain some confidence we will move it voting 16:09:36 sripriya: still working thru' them .. I'm hoping existing gate-functional-dsvm hook should be sufficient but it still to be seen 16:10:06 Next on Tokyo summit .. 16:10:34 sridhar_ram: cool 16:10:55 I've grabbed a BoF (Birds-of-Feather) session for OpenStack NFV Orchestration ... 16:11:20 it will be on Tuesday 12:05 - 12:40 ... so please mark your calendars 16:11:37 it will show up in the main conference schedule soon 16:12:16 It is a fish bowl style session (upto 30 people I guess) .. 16:12:37 we should invite our partner orgs like ETSI, TOSCA, OPNFV and other developers to attend and provide inputs 16:13:00 Last .. our Liberty Release deadline in next week Sept 25 ! 16:13:05 just one week to go .. 16:13:12 lets move on.. 16:13:31 #topic Liberty Release Status 16:13:42 #topic Health Monitoring 16:13:49 bobh: please take over ... 16:14:04 I pushed the basic framework last night as WIP 16:14:19 bobh: that's cool ! 16:14:34 I'm still working on the link between the template processing in heat.py and the monitoring thread 16:15:00 Hoping to have that in place tomorrow so I can start testing 16:15:43 Working with tbh for http-ping, unit tests and devrefs 16:15:44 bobh: that's nice .. 16:16:18 I think it will be similar enough to the ETSI spec and we can tweak from that 16:17:11 There are some larger design issues that will need to be addressed at some point but this is a good start 16:17:44 bobh: agree, the last iteration of the spec looks reasoanable .. 16:17:59 That's all I have unless anyone has questions/comments 16:18:31 bobh: of course we need to get the spec merged ... ;-) 16:18:41 i guess we have 2 WIP for monitoring framework.. one from bobh and other from tbh 16:18:45 I know we doing are things in parallel .. as we are short on time 16:18:49 minor details... I'll push the final(?) version today 16:19:37 bobh: I'm quite happy w/ the spec.. one thing I would like to see is the abstract class interface .. we can take that up in the code-review 16:20:08 I copied the class from Mgmt but we can modify as needed 16:20:34 Its definitely more than we need at the moment but I think it provides a good framework for future expansion 16:20:59 bobh: agree.. 16:21:21 team - if you've any last minute thoughts on health-mon .. please review https://review.openstack.org/#/c/202126/ 16:21:47 bobh: anything else ? 16:22:00 not that I can think of 16:22:32 sripriya: I believe tbh and bobh are consolidating the two WIP patches... 16:22:39 tbh: ping 16:23:24 okay .. lets move on to SFC 16:23:28 sridhar_ram: i see 16:23:33 s3wong: ping 16:23:46 #topic Tacker-SFC 16:24:15 trozet: hi, can you provide some quick updates from OPNFV-SFC of things..? 16:24:27 sridhar_ram: sure 16:24:33 it will be good to hear where we stand w.r.t SFC & Tacker ! 16:24:54 so OPNFV has a scope limitation where we cannot release any VNFM/MANO application as part of a release 16:25:09 due to a conflict of vendor interests when OPNFV was formed 16:25:32 hello, sorry 16:25:54 so this is kind of a dumb limitation, and the chair of OPNFV (Chris Price) has agreed to re-examine the scope after our next release which is Feb 2nd 16:26:25 so the tl;dr is they wont let Tacker be in the next release as an official part 16:26:55 but I also lead an installer project in OPNFV (Apex), I will continue to do the Tacker work with you guys upstream and then get it into our installer for the next release 16:27:06 as an unofficial "experimental testing feature" 16:27:44 trozet: that's awesome! 16:28:29 to make that happen.. which openstack release would OPNFV release base on ? 16:28:37 s3wong: hi 16:28:54 it will be Liberty O/S and Lithium ODL 16:29:43 trozet: I guess we need to find a way to get the tacker SFC work done off our liberty branch 16:30:15 sridhar_ram: is that OK with you? 16:30:51 overall I'm fine w/ tacker not directly part of OPNFV release .. and going in as experimental feat on top of B will give us room to iron out our initial issues.. 16:31:28 ok 16:31:54 trozet: that's a question to the team here ... looking for ideas how to support this effort 16:32:18 with the constraint that OPNFV B release is going to be Liberty based 16:32:21 team - any thoughts ? 16:32:27 yeah I will probably need some help as this is the first time I've modified/done any openstack dev work 16:32:54 sounds like a good plan - a head start on Liberty will be nice 16:33:07 trozet: don't worry, we have many folks here to guide / help you 16:34:06 so I will work on the spec, shall that be proposed on master or a liberty branch? 16:34:10 bobh: the challenge is to have make a liberty release with everything else we have - SFC and then somehow accommodate SFC 16:34:37 sridhar_ram, trozet: when is OPNFV B release going to be? 16:34:49 s3wong: set for Feb 2nd 16:34:58 one option is to make multiple Liberty release - kinda along the lines of ODL's SR1, SR2... 16:36:24 sridhar_ram: I guess a special Liberty Tacker release for SFC is probably the answer 16:37:01 sridhar_ram: I don't think as a team we want to get into a habit of having in-cycle release, probably just special case 16:37:23 do we work on tacker master for SFC and then backport to Liberty for a SR once OpenStack is released for Liberty...? 16:37:26 s3wong: agree, this would be an exception.. 16:38:01 sridhar_ram: since Tacker isn't an OpenStack tent project, and even if it is, it would likely forever be 'release: independent', one option would actually be releasing Liberty sometimes in Dec/Jan with the SFC stuff integrated 16:38:03 s3wong: we need better planning w/ downstream projects .. and even other way around 16:38:35 here is a proposal :) 16:39:15 lets pull a stable/release close to 09/25 for features currently in the pipeline.. 16:39:31 * stable/liberty 16:40:01 pull a sfc branch off stable/liberty and have trozet develop code there ... do another liberty-sfc release 16:40:21 ... after merge to stable/liberty 16:40:41 then double commit the sfc feature to master .. towards mitaka 16:41:11 makes sense to me 16:41:14 sridhar_ram: agree on the plan 16:41:25 sridhar_ram: +1 16:42:12 s3wong: you get the last word ! 16:42:53 there is bit of branch merge logistics .. but doable IMO 16:43:01 sridhar_ram: I am OK with that --- given that based on this cycle's development propensity, I don't think we expect a lot of checkins between summit and Jan/Feb (when I presume the SFC stuff would have to work very well) 16:43:36 so presumably there should be minimal syncing between the stable branch and the sfc-dev branch 16:43:51 s3wong: agree... 16:44:35 I hope sfc branch is a short lived one and we quickly get things into master to march on 16:45:15 I've seen the speed at which trozet codes - so I'm not worried ;-) 16:45:15 sridhar_ram: so you are going to take the action item to pull the stable and sfc-dev branches? :-) 16:45:27 just a heads up also - there is development work to be done in the lower layers (OVS, ODL) which will be handled in the OPNFV SFC project 16:45:52 s3wong: yeah, I'll take up the branch logistics 16:45:53 so the main tacker sfc stuff working with ODL will depend on ^ being complete 16:46:24 trozet: is the OPNFV team pushing to get NSH support (finally) into OVS upstream? 16:46:33 trozet: understood, there are many moving parts here... 16:47:04 s3wong: its complicated :) Intel is pushing a patch upstream (they estimated within the next 1-2 weeks) that allows Ethernet+NSH, but only with dpdk 16:47:31 trozet: Oh, that is funky... 16:47:32 s3wong: OPNFV SFC wants a dpdk and non-dpdk solution 16:48:26 s3wong: using the unofficial patch nshv8 that Cisco made works with ODL SFC, but it doesn't work when integrating into openstack without hacks and workarounds 16:48:53 s3wong: so there are problems to solve and work to be done there, before we can actually render a service function chain into OVS from Tacker 16:49:12 trozet: with Lithium already released .. is this even doable ? that is to get openstack --> ODL-Li --> patched OVS-NSH to work ? 16:49:39 trozet: yeah, I remember seeing the patch from the Cisco (vCider folks, I believe) on ovs-dev back in early 2013; at that time blp's point is we need to wait for official UDP port number for NSH, and now we have had it for about a year now, and still not in OVS :-) 16:50:10 * sridhar_ram notes, we are 10mins mark 16:50:15 sridhar_ram: well how it goes is we have to release by default with Liberty and Lithium...however our installer can have a flag to install a certain version of ODL 16:50:52 trozet: yeah, we at Neutron networking-sfc is also waiting for it. Neutron leadership does NOT allow OVS forks to be the data path backend 16:51:00 sridhar_ram: so we have some flexibility there, especially with calling Tacker SFC experimental 16:51:06 I see... 16:51:39 however the openstack pieces will be Liberty .. is there any room to pick something newer ? 16:52:01 s3wong: so you are working on the networking-sfc? We can work together to make sure the spec is abstract enough to meet both versions of SFC (odl + neutron) 16:52:18 trozet: yes, I am a core there as well 16:52:51 sridhar_ram: technically yes, we could have a flag to install a different openstack, but our limitation is our installer (RDO-Manager) which relies on RDO 16:53:02 trozet: and we should work together to get Tacker to work with both 16:53:13 sridhar_ram: I'm not sure when RDO would have M support, and if Tacker is included 16:53:37 trozet: okay, then the proposal we painted earlier is the best IMO.. 16:53:42 i think so 16:54:33 s3wong: trozet: I do believe Tacker-SFC could *eventually* converge with networking-sfc... but for now lets not make networkign-sfc as a dependency for Tacker-SFC 16:54:49 we have enough dependency already - ODL, OVS, etc... :) 16:54:52 oh ok you want to keep it separate 16:55:07 sridhar_ram: I don't think for Liberty we are making it a dependency 16:55:10 but totally agree and support the common cause to get OVS pieces in plance 16:55:12 *place 16:55:42 sridhar_ram, trozet: but my guess is eventually even ODL SFC team would want to integrate with networking-sfc 16:55:56 s3wong: agree.. 16:56:10 here is the slide deck I circulated to OPNFV-SFC team.. 16:56:12 #link https://drive.google.com/open?id=18AGaiysVgHOd_fIHVpObMO7zUkMjJGOQ98CUwkxU1xo 16:56:15 s3wong: that party is still fuzzy to me as well. 16:56:36 it show the different phases on how we could evolve Tacker SFC towards networking-sfc 16:56:45 sridhar_ram, s3wong: but for Tacker, you want one SFC API that serves both ODL-SFC and networking-sfc, no? 16:57:04 trozet: that would be ideal 16:57:08 trozet: absolutely 16:57:21 atleast that should be the goal.. 16:57:22 sridhar_ram: ok just wanted to make sure that was the intent 16:58:10 NFV operators wouldn't (or shouldn't ) care how a SF chain is realized underneath 16:58:25 folks we are close to our time limit.. 16:58:28 lets wrap.. 16:58:52 lets march on w/ the plan we proposed earlier... 16:59:08 I'll take the action items to do the branch logistics.. 16:59:25 meanwhile lets get the health-mon in place by next week... 16:59:33 reviewers - please help out! 16:59:59 going to be busy week.. 17:00:05 talk to all next week ! 17:00:11 #endmeeting